Radiologic Physics Taught Through Cases

Jonathon A. Nye, PhD

Associate Professor
Department of Radiology and Imaging Sciences
Emory University School of Medicine
Atlanta, Georgia

231 illustrations

Thieme
New York • Stuttgart • Delhi • Rio de Janeiro
Library of Congress Cataloging-in-Publication Data
is available with the publisher
© 2020. Thieme. All rights reserved.
Thieme Publishers New York
333 Seventh Avenue, New York, NY 10001 USA
+1 800 782 3488, customerservice@thieme.com
Georg Thieme Verlag KG
Rüdigerstrasse 14, 70469 Stuttgart, Germany
+49 [0]711 8931 421, customerservice@thieme.de
Thieme Publishers Delhi
A-12, Second Floor, Sector-2, Noida-201301
Uttar Pradesh, India
+91 120 45 566 00, customerservice@thieme.in
Thieme Publishers Rio de Janeiro,
Thieme Publicações Ltda.
Edifício Rodolpho de Paoli, 25º andar
Av. Nilo Peçanha, 50 – Sala 2508
Rio de Janeiro 20020-906 Brasil
+55 21 3172 2297
Cover design: Thieme Publishing Group
Typesetting by DiTech Process Solutions, India
Printed in USA by King Printing Company, Inc.
ISBN 978-1-62623-967-8
Also available as an e-book:
eISBN 978-1-626-23971-5
Important note: Medicine is an ever-changing science
undergoing continual development. Research and clinical
experience are continually expanding our knowledge, in
particular our knowledge of proper treatment and drug
therapy. Insofar as this book mentions any dosage or appli-
cation, readers may rest assured that the authors, editors,
and publishers have made every effort to ensure that such
references are in accordance with the state of knowledge at
the time of production of the book.
Nevertheless, this does not involve, imply, or express any
guarantee or responsibility on the part of the publishers in
respect to any dosage instructions and forms of applications
stated in the book. Every user is requested to examine
carefully the manufacturers’ leaflets accompanying each
drug and to check, if necessary in consultation with a phy-
sician or specialist, whether the dosage schedules mentioned
therein or the contraindications stated by the manufacturers
differ from the statements made in the present book. Such
examination is particularly important with drugs that are
either rarely used or have been newly released on the market.
Every dosage schedule or every form of application used is
entirely at the user’s own risk and responsibility. The authors
and publishers request every user to report to the publishers
any discrepancies or inaccuracies noticed. If errors in this
work are found after publication, errata will be posted at
www.thieme.com on the product description page.
Some of the product names, patents, and registered
designs referred to in this book are in fact registered trade-
marks or proprietary names even though specific reference
to this fact is not always made in the text. Therefore, the
appearance of a name without designation as proprietary is
not to be construed as a representation by the publisher that
it is in the public domain.
This book, including all parts thereof, is legally protected by
copyright. Any use, exploitation, or commercialization out-
side the narrow limits set by copyright legislation without the
54321
publisher’s consent is illegal and liable to prosecution. This
applies in particular to photostat reproduction, copying,
mimeographing or duplication of any kind, translating, prep-
aration of microfilms, and electronic data processing and
storage.
Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x

1. Fluoroscopy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Rebecca Milman Marsh and Michael Silosky

1.1 Case 1: SID, ABC, and Radiation 1.7 Case 7: Digital Subtraction
Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Angiography and Motion
Artifacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2 Case 2: Reference Air Kerma and
1.8 Case 8: Fluoroscopy Modes and
Skin Dose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Dose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.3 Case 3: Collimation . . . . . . . . . . . . . . . . . . . 6
1.9 Case 9: Equalization Filters . . . . . . . . . . 18

1.4 Case 4: Anti-scatter Grids . . . . . . . . . . . . 8 1.10 Case 10: Cone Beam Computed
Tomography . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.5 Case 5: Patient Shielding . . . . . . . . . . . . 10
1.11 Review Questions . . . . . . . . . . . . . . . . . . . . 22
1.6 Case 6: CT Fluoroscopy . . . . . . . . . . . . . . 12

2. Mammography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Ingrid S. Reiser

2.1 Case 1: Magnification Imaging. . . . . . 26 2.7 Case 7: Microcalcification-like

Appearance Caused by a Detector
2.2 Case 2: Focal Spot Size Selection Artifact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
in Magnification Views . . . . . . . . . . . . . . 27
2.8 Case 8: Artifact due to Imperfection
2.3 Case 3: X-ray Acquisition Technique in Compression Paddle . . . . . . . . . . . . . . 37
Factors in Mammography . . . . . . . . . . . 29
2.9 Case 9: Patient Motion Causing
2.4 Case 4: Digital Breast Tomosynthesis: Blurred Parenchymal Structure
Artifacts due to High-Contrast in a Mammogram. . . . . . . . . . . . . . . . . . . . 38
Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 2.10 Case 10: EMI Artifact due to LVAD
2.5 Case 5: Effect of Image Device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Post-Processing on the Appearance 2.11 Review Questions . . . . . . . . . . . . . . . . . . . . 40
of a Mammogram . . . . . . . . . . . . . . . . . . . 32

2.6 Case 6: Artifact due to Detector

Row Dropout . . . . . . . . . . . . . . . . . . . . . . . . . 35
 Contents

3. Computed Tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Karen L. Brown and Jason R. Gold

3.1 Case 1: Ring Artifact . . . . . . . . . . . . . . . . . 44 3.6 Case 6: Displayed Volume CT Dose

Index and Patient Size . . . . . . . . . . . . . . . . 51
3.2 Case 2: Effect of Patient Size
3.7 Case 7: Beam Hardening Artifact . . . 52
on CT Number Accuracy . . . . . . . . . . . . . 45
3.8 Case 8: Partial Volume Artifact . . . . . 53
3.3 Case 3: Effect of kV Selection
3.9 Case 9: Metal Artifact . . . . . . . . . . . . . . . 54
on Image Quality and Dose . . . . . . . . . 46
3.10 Case 10: Motion Artifact . . . . . . . . . . . . 55
3.4 Case 4: Image Quality Variation
with Reconstructed Slice 3.11 Review Questions . . . . . . . . . . . . . . . . . . . . 56
Thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

3.5 Case 5: Image Quality Variation

with Reconstruction Filter . . . . . . . . . . 49

4. Magnetic Resonance Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

Puneet Sharma

4.1 Case 1: Appearance of Discrete 4.6 Case 6: Application of Fat-Suppressed

Image Ghosts on Abdominal Sequences in the Pelvis did not
Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 Reveal the Expected Contrast . . . . . . . . 71

4.2 Case 2: A Well-Defined Area of 4.7 Case 7: T1-Weighted Gradient

Signal Hyperintensity Appears Echo of the Abdomen Shows
Bilaterally at the Level of the Marked Artifact Medially on Both
Internal Auditory Canal on Coronal and Axial FOV, Obscuring
Diffusion-Weighted MRI, Affecting Visualization of Soft Tissues . . . . . . . . 74
Visualization of Surrounding
4.8 Case 8: Abnormal Dark Fluid Seen
Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
in the Bladder of an Axial Single-Shot
T2-Weighted Sequence, but not on
4.3 Case 3: Appearance of Extra
Location-Matched 3D T2
Field-of-view Anatomy on the
Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Inferior Portion of Sagittal 3D
T2-Weighted Acquisition of the 4.9 Case 9: Postcontrast T1-Weighted
Spine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 Gradient Echo Reveals Patchy
Enhancement in the Anterior
4.4 Case 4: Precontrast, Axial 3D
Septal Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
T1-Weighted Gradient Echo with
Fat Suppression Shows Adequate 4.10 Case 10: Significant Signal-to-Noise
Anatomical Detail, but Minor Edge Variation Across the FOV, Creating
Ripple and Blur that is Presumed Nondiagnostic Image Quality . . . . . . . 82
to be Motion . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.11 Review Questions . . . . . . . . . . . . . . . . . . . . 85
4.5 Case 5: Dark Etching Appears
at the Boundary of Fat and
Soft-Tissue Layers . . . . . . . . . . . . . . . . . . . . 69

vi
 Contents

5. Nuclear Medicine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Jonathon A. Nye, James R. Galt, and John N. Aarsvold

5.1 Case 1: Degraded Resolution of a 5.6 Case 6: The Effects of Image

Whole-Body Planar 99mTc Methylene Di- Smoothing . . . . . . . . . . . . . . . . . . . . . . . . . . 104
phosphonate Image . . . . . . . . . . . . . . . . . . . 91
5.7 Case 7: Choosing the Correct
5.2 Case 2: Effect of Positron Range Acquisition Image Matrix Size . . . . . 106
on Image Quality and Resolution . . . 94
5.8 Case 8: Assessing Patient Motion
in Myocardial Perfusion Imaging . . 108
5.3 Case 3: Standardized Uptake
Value in Positron Emission 5.9 Case 9: Bremsstrahlung Imaging
Tomography (Noise Bias) . . . . . . . . . . . . 97 of 90Y Microspheres Liver
Embolization . . . . . . . . . . . . . . . . . . . . . . . . 110
5.4 Case 4: The Impact of Attenuation
Correction in PET . . . . . . . . . . . . . . . . . . . . 99 5.10 Case 10: Degraded Image
Quality from an Improper
5.5 Case 5: Iterative Reconstruction
Collimator . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
and Choosing the Number of
Iterations and Subsets . . . . . . . . . . . . . . 101 5.11 Review Questions . . . . . . . . . . . . . . . . . . . 114

6. Ultrasound Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

Zheng Feng Lu

6.1 Case 1: Pulse-Echo Imaging 6.6 Case 6: Range Ambiguity in

Principle and Speed of Sound B-Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Propagation . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.7 Case 7: Shadowing and
6.2 Case 2: Array Transducers Enhancement (Increased Through
and Sound Frequency . . . . . . . . . . . . . . 120 Transmission) . . . . . . . . . . . . . . . . . . . . . . . 130

6.8 Case 8: Harmonic Imaging . . . . . . . . . 132

6.3 Case 3: Nonuniformity
(Array Transducer Element
6.9 Case 9: Ultrasound Image
Dropouts) . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
Display on Scanners and
6.4 Case 4: Pulse-Echo Imaging in Reading Rooms. . . . . . . . . . . . . . . . . . . 133
Acquisition Controls . . . . . . . . . . . . . . . . 124
6.10 Case 10: Doppler Ultrasound
6.5 Case 5: Reflection (Boundary Aliasing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Conditions)—Reverberation
6.11 Review Questions . . . . . . . . . . . . . . . . . . . 138
Artifacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

vii
 Contents

7. Image Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

Jonathon A. Nye and Randahl C. Palmer

7.1 Case 1: Filtering and Edge 7.3 Case 3: Fused Image Display
Enhancement . . . . . . . . . . . . . . . . . . . . . . . 144 of Multiple Modalities . . . . . . . . . . . . . . 148

7.2 Case 2: Maximum Intensity 7.4 Case 4: Multimodality Image

Projection . . . . . . . . . . . . . . . . . . . . . . . . . . . 146 Registration . . . . . . . . . . . . . . . . . . . . . . . . . 150

Answer Key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

viii
Preface
Radiology residents gather their medical physics
knowledge from multiple sources, often beginning
with their first encounter with a radiologic image.
Although many educational approaches start with
fundamental physical concepts and work toward
image generation, they are likely to require an
extended period of time to build a conceptual frame-
work. Arguably, the clinical demands of residency
training do not always allow for a traditional class-
room approach, as it can be much more efficient to
learn about radiologic imaging principles during the
course of a clinical rotation. Therefore, a hybrid
approach may be more amenable, which begins with
a study of images commonly encountered during
diagnostic radiology training and provides a straight-
forward and compact explanation of the physical
factors underlying the creation and displayed con-
trast of these images. To that end, this book presents a
number of common physical concepts in diagnostic
radiology, which may be encountered by a resident at
their clinical workstation. Although the breadth of
this topic area is large, the goal of this text is to provide
examples relevant to diagnostic radiology training,
thereby proving to be of high value to the learner.
Chapters are divided according to modality, each
having 10 topics presented in a case format that is
meant to quickly convey information with an image
followed by a brief explanation. Some important
topics, such as radiation safety, do not lend them-
selves to teaching from images but are part of impor-
tant encounters like occupational or patient hazards
during a fluoroscopic-guided procedure. In these
cases, schematics are provided to assist in teaching.
The reader is encouraged to consult the chapter refer-
ences for further discussion. Review questions are
provided at the end of each chapter to reinforce the
case concepts.
Jonathon A. Nye, PhD
ix
 Contributors
John N. Aarsvold, PhD Jonathon A. Nye, PhD
Associate Professor Associate Professor
Department of Radiology and Imaging Sciences Department of Radiology and Imaging Sciences
Emory University School of Medicine Emory University School of Medicine
Atlanta, Georgia Atlanta, Georgia

Karen L. Brown, MHP, CHP, DABR Randahl C. Palmer, MS

Diagnostic Imaging Physicist Resident,
Department of Radiology Department of Radiology and Imaging Sciences
Penn State College of Medicine Emory University School of Medicine
Hershey, Pennsylvania Atlanta, Georgia

James R. Galt, PhD Ingrid S. Reiser, PhD

Professor Associate Professor
Department of Radiology and Imaging Sciences Department of Radiology
Emory University School of Medicine The University of Chicago
Atlanta, Georgia Chicago, Illinois

Jason R. Gold, DO Puneet Sharma, PhD

Resident Assistant Professor
Department of Radiology Department of Radiology and Imaging Sciences
Penn State Milton S. Hershey Medical Center Emory University School of Medicine
Hershey, Pennsylvania Atlanta, Georgia

Zheng Feng Lu, PhD Michael Silosky, MS

Professor Assistant Professor
Department of Radiology Department of Radiology
The University of Chicago University of Colorado School of Medicine
Chicago, Illinois Aurora, Colorado

Rebecca Milman Marsh, PhD

Associate Professor
Department of Radiology
University of Colorado School of Medicine
Aurora, Colorado

x
| 09.11.19 - 01:52
1 Fluoroscopy
Rebecca Milman Marsh and Michael Silosky
Introduction
Shortly after the discovery of X-rays in the late
nineteenth century, fluoroscopy was developed to
enable visualization of moving anatomy. Today,
fluoroscopy is used for diagnosis and guidance of
clinical procedures. Applications of fluoroscopy
are found throughout medicine, including radi-
ology, cardiology, urology, and speech pathology.
Fluoroscopy has the same tradeoffs between
image quality and radiation exposure as other
X-ray-based modalities. During a fluoroscopy
exam, multiple factors affect radiation dose to
patients and staff. Some may be directly controlled
by the fluoroscopy operator while others are
dictated by patient habitus and the procedure
being performed. Some interventional procedures
can be very complex, requiring long fluoroscopy
times to accomplish the clinical task.
Fluoroscopy differs from most other imaging
modalities in that the physician is often directly
involved in image acquisition. In addition to
ensuring that image quality is adequate to achieve
the clinical task being performed, the operator is
responsible for dose management throughout the
exam, not only for the safety of the patient but also
for all staff members present. Understanding fluo-
roscopy imaging parameters and their effect on
both image quality and dose is essential for anyone
operating or supervising the operation of a fluoro-
scopy system.
An essential function of modern fluoroscopy
systems is the continuous adjustment of X-ray
tube output to maintain adequate image quality
throughout the procedure. This feedback loop
between the detection system and the X-ray tube
is commonly referred to as automatic brightness
control (ABC), referring to the goal of maintaining
a constant brightness on the output phosphor of
the image intensifier.1 With the advent of digital
detectors, this same concept is sometimes
referred to as automatic dose control (ADC).
Throughout this chapter, we will use the tradi-
tional term ABC.
1
| 09.11.19 - 01:52

Fluoroscopy

1.1 Case 1: SID, ABC, and Table 1.1 Acquisition parameters, reference air kerma
rates, and estimates of skin dose are shown for a range
Radiation Output of source-to-image receptor distances
Reference Skin dose
1.1.1 Background SID AKR rate
kV mA
● A patient undergoes a fluoroscopy-guided L1 (cm) (mGy/ (mGy/
minute) minute)
kyphoplasty procedure to treat vertebral body
119 65 5.1 3.03 5.6
compression.
110 65 4.4 2.62 4.79
● At the start of the procedure, the operator places
100 66 3.3 2.07 3.89
the image detector 10 cm above the patient.
90 67 2.6 1.65 3.18
● In order to properly insert the cannula into the
Abbreviations: AKR, air kerma rate; SID, source-to-image
vertebral body, the operator has to raise the receptor distance.
detector to 30 cm above the patient.
In addition, the distance between the X-ray source
1.1.2 Findings and the patient’s skin (known as the source-to-
skin distance [SSD]) directly impacts
Increasing the distance between the X-ray source patient dose. Dictated by the inverse square law,
and the image detector increased the reported air doubling the distance between the radiation
kerma rate (AKR) by a factor of approximately 1.5. source and an object will result in a reduction
in exposure to that object by 75%. In the case
1.1.3 Discussion described here, the increase in X-ray tube
output can be calculated using the equation:
Patient dose is dependent on the source-to-image
receptor distance (SID) which is the distance 1
between the source of X-rays (the focal spot of the ðr2 =r1 Þ2
X-ray tube) and the image receptor. As SID
changes, the fluoroscopy system adjusts acquisi- Thus, where r1 = 50 cm and r2 = 100 cm.
tion parameters, directly affecting the radiation 1 1
¼
output. As discussed in the beginning of this ð100=50Þ2 4
chapter, the ABC algorithms used by fluoroscopy
systems adjust tube output to maintain image or 25%.
appearance. For a fixed X-ray output, as the When determining how the patient and equip-
distance to the image receptor increases, the ment should be positioned for an exam, one
amount of radiation that reaches the image should consider the location(s) where the operator
receptor decreases. For example, at an SID of needs to directly access the patient, how much
100 cm, only 25% as many X-ray photons will working space is needed between the patient and
intercept the detector compared to an SID of the detector, any special positioning needs of the
50 cm. To compensate for fewer photons reaching patient, and operator ergonomic factors such as
the receptor at larger SIDs, ABC algorithms table height. While some of these factors are
increase X-ray tube output. Typically, this is done dictated by the needs of the specific exam and
by changing one or more of the following parame- operator, ensuring that SID is as small as practical
ters: kV, mA, pulse width, or beam filtration. This can help to reduce patient dose.
will result in an increase in reference air kerma
(AK) and, assuming a constant distance between
the X-ray source and the patient’s skin, an increase
1.1.4 Resolution
in patient dose. This is illustrated in ▶ Table 1.1 The following steps can help reduce patient dose
where entrance dose to a phantom was measured while maintaining affecting image quality. First,
under conditions of variable SID. for fluoroscopic systems with fixed SID, as is the

2
| 09.11.19 - 01:52
case with many mobile C-arms, the patient should
be positioned as close to the image receptor (as far
from the source) as is practical, given the needs of
the procedure. Second, for fluoroscopic systems
with variable SID, such as those used in interven-
tional radiology or cardiac catheterization labs,
1.1 Case 1: SID, ABC, and Radiation Output
patient positioning may be primarily dictated by
the procedure being performed. Once the patient
has been positioned and the fluoroscopy system is
oriented as desired, the SID should be reduced as
much as is practical by moving the image receptor
toward the source.
3
| 09.11.19 - 01:52
Fluoroscopy
1.2 Case 2: Reference Air Kerma
and Skin Dose
1.2.1 Background
● A patient undergoes a superior mesenteric arte-
riogram for embolization of a pseudoaneurysm
in the transverse colon.
● At the end of the exam, the total reference AK is
5100 mGy.
● The case is referred to Radiation Safety so that a
peak skin dose (PSD) estimate can be performed.
1.2.2 Findings
The estimated PSD is 7000 mGy, approximately
1.5 times the reference AK displayed by the fluoro-
scopy system.
1.2.3 Discussion
Fluoroscopy operators can monitor the use of radia-
tion during procedures by paying attention to
machine-reported dose metrics. All modern fluoro-
scopy systems are required to display the reference
AK and reference AKR, where the reference AK is
equivalent to the dose to air at a specific reference
point. (The exact location of the reference point
varies based on equipment vendor and model.)
These values provide the operator with real-time
dosimetry information throughout a case. Conse-
quently, it is important to know how these values
are measured and how they relate to patient expo-
sure and patient skin dose, specifically.
For fluoroscopy procedures, the primary radia-
tion safety concern for patients is the PSD, which
is the maximum dose to any single area of the skin.
4
It is notable that in this case, the estimated PSD is
substantially greater than the AK reported by the
machine. Patient skin dose is affected by several
factors including table attenuation, backscatter,
differences in how dose is deposited in various
materials, and patient positioning. None of these
factors are considered when the system calculates
the displayed AK. While a thorough discussion is
beyond the scope of this text, special attention
should be paid to the effects of patient positioning,
specifically SSD.
The most common reason for large differences
between the displayed AK and the actual skin dose
is patient size. In the case described here, the pa-
tient was morbidly obese, and the location where
the X-ray beam entered the patient’s skin was much
closer to the X-ray source than the location where
the fluoroscopy system calculated the AK. As a
consequence, the AK at the patient’s skin was
greater than the reference AK. As the AK at the
patient’s skin increases, skin dose will also increase.
These concepts are illustrated in ▶ Fig. 1.1. The
X-ray tube is under the patient table with the X-ray
focal spot (the source of the X-ray beam) indicated
by a white “x.” The AK reference point (located
65 cm from the focal spot) is indicated by a black
“x.” In ▶ Fig. 1.1a, the patient is closer to the X-ray
tube than the AK reference point is. Here the
entrance AK at the patient’s skin will be greater than
the reference AK. The opposite happens in
▶ Fig. 1.1b, where the patient is farther from the
source than the reference point. In this case,
entrance AK will be lower than the reference AK. It
should be noted that variation between the
displayed AK and the entrance AK is governed
by the inverse square law—as an object moves
farther away from the source of radiation, the expo-
sure decreases as a factor of the square of the
Fig. 1.1 The relationship between
the X-ray focal spot (white “x”), the air
kerma reference point (black “x, and
the entrance point of the patient’s skin
are shown. (a) In this image the
patient’s skin is closer to the X-ray
source than is the air kerma reference
point, meaning that the reference air
kerma displayed by the system will
underestimate the patient’s skin dose.
(b) Here the patient’s skin is further
from the X-ray source. Consequently,
the air kerma displayed by the system
will overestimate the patient’s skin
dose.
| 09.11.19 - 01:52
distance. Consequently, entrance AK and actual
patient skin dose may be larger or smaller than the
reference AK.
The patient’s position relative to the AK refer-
ence point is not the only factor that affects the
PSD. If, for example, the C-arm is rotated during
the fluoroscopy exam, the dose will be “spread
out” over different areas of the patient’s skin.
Similarly, if the table or C-arm is translated so
that the imaging field of view moves from the
groin to the chest, the total AK displayed by the
system will include radiation delivered to
different portions of the skin. While any detailed
PSD estimate should be performed by a qualified
medical physicist, understanding the relationship
between the displayed AK and patient skin dose
and recognizing when these values are likely to
1.2 Case 2: Reference Air Kerma and Skin Dose
diverge will allow the fluoroscopy operator to
manage radiation dose during a procedure.
1.2.4 Resolution
If the AK reference point is at 65 cm from the focal
spot, then the AK at 53 cm from the focal spot is
(65/53)2 or 1.5. This means that if the displayed AK
is 5100 mGy, then the patient entrance AK will be
approximately 7700 mGy. Other factors, such as
attenuation of the X-ray beam by the table, back-
scatter generated by the patient, and a conversion of
entrance skin exposure to absorbed skin dose must
be considered to determine the PSD. However,
entrance AK can be used as a reasonable surrogate
for PSD when considering radiation management
during a fluoroscopy procedure.
5
| 09.11.19 - 01:52
Fluoroscopy
1.3 Case 3: Collimation
1.3.1 Background
● Patient underwent a fluoroscopy-guided
exchange of a retrograde left nephroureteral tube.
● Two digital spot images were acquired during
the procedure with different amounts of physical
collimation.
1.3.2 Findings
● Collimation was used to reduce the field of view
(from 21.9 × 28.4 cm to 16.5 × 19.9 cm; ▶ Table 1.2).
● Dose area product (DAP) was reduced by 47%.
1.3.3 Discussion
This case illustrates that collimation (also some-
times referred to as “coning in”) can reduce radia-
tion exposure to both patients and staff and
improve image quality. Collimation uses lead shut-
ters inside the X-ray tube housing to reduce the
imaging field of view. This enables one to image
(and irradiate) only the patient anatomy essential
for performing the procedure.
Fig. 1.2 Digital spot images acquired without collimation (a) and with collimation (b).
6
Table 1.2 The collimated field size and dose metrics for
the uncollimated and collimated images shown in
▶ Fig. 1.2
Uncollimated Collimated
(left) (right)
Collimated field 21.9 × 28.4 cm 16.5 × 19.9 cm
size (620 cm2) (328 cm2)
DAP (mGycm2) 446 278
AK (mGy) 1.3 1.3
Changes in collimation will not substantially affect
AK. However, collimation will affect DAP (some-
times called Kerma Area Product, or KAP) which is
another dose metric that is commonly displayed on
fluoroscopy systems. DAP is the product of AK and
field size, so even if AK remains constant, DAP will
change proportionally to the change in X-ray field
size. For procedures using single fields (i.e., the
position of the X-ray tube remains constant relative
to the patient throughout the exam), this will not
reduce the PSD substantially but will reduce the
effective dose to the patient. For procedures with
multiple fields, proper collimation can reduce the
likelihood of having overlapping radiation fields
and potentially reduce PSD. In addition, collimation
can greatly reduce staff exposure since a smaller
imaging field will produce less scatter radiation.
| 09.11.19 - 01:52
Finally, proper collimation has the additional
benefit of enhanced image quality. Collimated fields
can be used to exclude very high- or low-attenu-
ating tissues, reducing the variation in signal inten-
sity across the field of view, which may improve
image contrast. Moreover, since proper collimation
reduces the amount of scatter created by patient
tissue, images will appear with improved contrast
(Table 1.2).
1.3 Case 3: Collimation
1.3.4 Resolution
In the example shown in ▶ Fig. 1.2, the field of
view was reduced from 21.9 × 28.4 cm (620 cm2)
in ▶ Fig. 1.2a to 16.5 × 19.9 cm (329 cm2) in
▶ Fig. 1.2b; a reduction of 47%. DAP also decreased
47%, i.e., from 807 mGycm2 (▶ Fig. 1.2a) to 428
mGycm2 (▶ Fig. 1.2a). AK was the same for both
images (1.3 mGy).
7
| 09.11.19 - 01:52
Fluoroscopy
1.4 Case 4: Anti-scatter Grids
1.4.1 Background
● A 5-year-old patient undergoes a fluoroscopy-
guided cardiac procedure.
● The patient had a similar procedure 6 months
prior where the total AK was 450 mGy.
● With a similar amount of fluoroscopy time and
the same number of exposures, the AK for the
current procedure was 1200 mGy.
1.4.2 Findings
● The same fluoroscopy system and imaging
protocol were used for both procedures.
● The images for the prior procedure (▶ Fig. 1.3a)
have poorer image quality than the images from
the later procedure (▶ Fig. 1.3b).
1.4.3 Discussion
X-ray photons that reach the image detector are
either primary or scattered photons. Primary
photons travel along a straight path through the
patient until they are absorbed by the detector.
Scattered photons are those that have interacted
with the patient’s tissue (or other objects) and
have changed direction from their original path.
Fig. 1.3 Fluoroscopy images of an anthropomorphic chest phantom without (a) and with (b) an anti-scatter grid.
8
Image quality is greatly influenced by the
scatter-to-primary ratio (SPR) or the ratio of the
amount of energy deposited in the detector by
scattered versus primary photons. If a fluoro-
scopy system has an SPR of 3, this indicates that
75% of the energy deposited in the detector is
from scattered photons while the remaining 25%
is from primary photons.
As the SPR increases, low-contrast objects are
more difficult to distinguish. Image quality can be
improved by using an anti-scatter grid. The grids
are made up of a series of septa made of lead or
other X-ray-attenuating material, and these septa
preferentially absorb scattered photons while
allowing more of the primary photons to pass
through. The grid is placed just in front of the
image receptor, decreasing the SPR, and improving
image quality. Since the grid also absorbs some of
the primary photons, the X-ray system must
compensate by producing more photons (i.e.,
increasing mAs), which increases patient dose. A
main determinant of how much scatter is
produced in a specific patient is patient habitus. As
the X-ray beam passes through additional tissue,
the number of scatter events increases. In other
words, thicker anatomy and larger patients will
produce more scatter than smaller patients.
Consequently, the anti-scatter grid has a substan-
tial effect on image quality when imaging adult
patients, but it has little effect when imaging
extremities or small patients.
| 09.11.19 - 01:52
1.4.4 Resolution
▶ Fig. 1.3 shows an image of an anthropomorphic
chest phantom. The image on the left (▶ Fig. 1.3a)
was obtained without an anti-scatter grid whereas
the image on the right (▶ Fig. 1.3b) was obtained
with the grid in place. The effect on image quality
is apparent by looking at the increase in detail
visible in the image on the right (obtained using
the grid.) However, for some clinical indications
the image quality in ▶ Fig. 1.3 may be sufficient.
The acquisition parameters for the two images are
also shown. The grid attenuates both primary and
1.4 Case 4: Anti-scatter Grids
scatter photons, causing the ABC system to
increase both kV and mA. As a result, the AKR
increased by a factor of almost 3.
Removing the anti-scatter grid is a common way
to reduce patient dose in fluoroscopy exams of
pediatric patients, where image quality is not
substantially affected. However, in most adult (or
adult-sized) patients, the grid is required to
achieve adequate image quality. It should be noted
that while many fixed fluoroscopy systems have
removable grids, most mobile C-arms have fixed
grids that cannot be removed.
9
| 09.11.19 - 01:52
Fluoroscopy
1.5 Case 5: Patient Shielding
1.5.1 Background
● A patient who is 30 weeks pregnant undergoes
a fluoroscopy-guided intervention to treat iliofe-
moral deep venous thrombosis.
● The patient is positioned prone on the table.
● A lead apron is placed under the patient’s
abdomen and pelvis (between the patient and
X-ray tube).
1.5.2 Findings
● The lead apron covers most of the imaging field
of view.
● Image quality is diminished as contrast is
reduced for important anatomical structures.
● The fluoroscopy system’s ABC algorithm
increases tube output to compensate for
additional attenuation caused by the presence
of the lead.
1.5.3 Discussion
Shielding materials (typically lead aprons) are
sometimes placed on or around patients with the
goal of reducing radiation exposure to staff. This
technique is also used with pregnant patients to
reduce fetal exposure to radiation. However,
placing lead under the patient may adversely
10
affects both image quality and dose. First, as
shown in ▶ Fig. 1.4, the presence of the lead apron
substantially reduces contrast throughout the
image. Second, since lead is highly attenuating, it
significantly affects the function of the ABC algo-
rithm increasing the X-ray tube output and radia-
tion dose to the patient, fluoroscopy operator, and
other staff present during the case.2
Fetal radiation exposure during an exam of the
mother varies substantially based on whether
the fetus is in or outside the imaging field of
view. If the fetus is in the imaging field of view, it
is exposed to the primary X-ray beam and radia-
tion dose could exceed 100 mGy, above which
there is an increased the risk of congenital
malformation, stillbirth, miscarriage, or mental
disability.3 If the fetus is outside of the imaging
field of view, the majority of radiation exposure
is from scatter generated within the mother and
the fetal dose is typically below 1 mGy.4 At such
low dose levels, there is no demonstrated
increased risk to the fetus. For a more detailed
discussion of management of pregnant patient
during fluoroscopy procedures, we refer the
reader to multisocietal guideline published by
Dauer et al.4
The use of patient shielding outside of the
imaging field of view has been advocated as a way
to reduce radiation dose to staff. However, it is
important to keep in mind that the vast majority
of operator exposure comes from scatter created
by interactions of the primary beam with patient
tissue. In other words, the radiation to which an
Fig. 1.4 A digital spot image of
a pregnant patient who had lead
apron placed between her skin and the
X-ray tube in an attempt to provide
additional protection to the fetus.
| 09.11.19 - 01:52
operator (and other staff) is exposed originates
from the patient. Consequently, shielding placed
on the patient can only provide a protective
benefit when an operator is standing in a few
specific locations. In addition, it has been demon-
strated that even under these conditions, the
protective benefit is negligible compared to that of
the operator’s own protective garments (i.e., lead
apron and glasses).5
Finally, due to the dynamic nature of fluoro-
scopically guided interventions, there is a risk that
shielding initially positioned outside of the
primary beam might end up in the imaging field of
1.5 Case 5: Patient Shielding
view. This carries the same risks as shielding
placed directly in the beam for patient protection,
i.e., reduced image quality and increased radiation
exposure to both the patient and staff.
1.5.4 Resolution
In summary, shielding materials should not be
placed on a patient. The benefit to the patient or
operator is negligible, but the risk of compromising
image quality is substantial. Furthermore, using
patient shielding may increase radiation dose to
everyone involved.
11
| 09.11.19 - 01:52
Fluoroscopy
1.6 Case 6: CT Fluoroscopy
1.6.1 Background
● An obese patient with diabetes and a history of
colon cancer presents with an enlarging right
lower lobe nodule with metabolic activity on a
recent positron emission tomography (PET)/
computed tomography (CT) scan.
● The patient undergoes a CT-guided right lung
mass biopsy (▶ Fig. 1.5).
● A total of 30 scans are performed. The total
CT dose index (CTDIvol) for the procedure is
1200 mGy. The dose length product (DLP) is
12,600 mGy-cm.
1.6.2 Findings
The patient was referred to radiation safety for
an estimate of PSD to determine if follow-up
evaluation for radiation-induced skin injury was
appropriate.
1.6.3 Discussion
While traditional fluoroscopy is based on planar
imaging techniques, CT fluoroscopy can be used
to create “real-time” cross-sectional images to
guide interventional procedures. Studies have
demonstrated that CT-guided biopsy is a low-risk
method of obtaining tissue samples of media-
stinal masses.6 Dose metrics used in CT differ
from those used in fluoroscopy. A basic under-
standing of CT dose metrics is a fundamental
component of dose management in CT-guided
interventional procedures.
The radiation output of a CT scanner is
commonly characterized by the CTDIvol and DLP.
CTDIvol is an estimate of dose, in mGy, to a
standard acrylic phantom. DLP is the product of
CTDIvol and the scan length and is reported in units
of mGy-cm. While these values can be used to
estimate patient effective dose, neither corre-
sponds well with skin dose, making the assess-
ment of risk for radiation-induced skin injuries
difficult. Studies have shown that skin dose can
range from approximately 49 to 65% of the
scanner-reported CTDIvol for scans performed
without table movement (as is often the scenario
during CT-guided interventional procedures).7 For
a CT scanner, the X-ray tube cannot be moved
closer or farther from the patient, meaning the
12
Fig. 1.5 A CT image obtained during the biopsy
procedure. The arrow indicates where the biopsy
needle is visible on the image.
X-ray tube will be closer to the patient’s skin if the
patient is obese, compared with the SSD for a
smaller patient. Consequently, if a patient is obese,
the ratio of PSD to CTDIvol may be even higher than
it would be for a smaller patient.
1.6.4 Resolution
In the case presented, the total CTDIvol is 1200
mGy and the DLP is 12,600 mGy-cm. These dose
metrics are considerably higher than those from a
typical diagnostic CT scan of the adult abdomen,
which has a CTDIvol of approximately 20 mGy and
a DLP of approximately 1000 mGy-cm.8 Using the
relationship cited above, where the PSD is approx-
imately 65% of the reported CTDIvol, PSD can be
conservatively approximated as 780 mGy. This is
well below the point at which radiation-induced
skin damage is expected.9 Of note, the patient in
the case presented here has diabetes, which, due
to the associated decrease in tissue perfusion, may
increase skin radiosensitivity. For a discussion of
other factors that can increase radiosensitivity,
please see the article by Balter et al.9
DLP is commonly used to calculate effective
dose.10 However, effective dose is not correlated
with PSD. While CTDIvol is a better indicator of
PSD, it suffers from some of the same limitations
as cumulative AK reported on a fluoroscopy
system, i.e., neither cumulative AK nor CTDIvol give
any indication of how the radiation exposure was
distributed across the patient’s skin. Although one
can generally assume that the radiation from a
| 09.11.19 - 01:52
CT-guided interventional procedure is evenly
distributed around the circumference of the
patient, when evaluating PSD, one should consider
whether all imaging was performed over a single
“slice” along the z-axis of the patient, or if the
patient was moved superiorly or inferiorly during
the procedure.
Another consideration when performing
CT-guided procedures is staff radiation exposure.
Similar to traditional fluoroscopy, staff will only be
exposed to the primary radiation beam if their
body enters the imaging field of view. On the other
1.6 Case 6: CT Fluoroscopy
hand, staff radiation exposure usually comes from
scatter created within the patient. Consequently,
when possible, staff should stand to the side of the
CT gantry, where the gantry provides a shielding
barrier between the patient and the staff member.
Alternatively, the operator can leave the room and
perform a single-slice CT scan. This can result in a
slightly higher radiation dose to the patient, but
the cumulative occupational dose savings to an
operator may be significant while the small
increase in radiation exposure may not pose any
additional risk to the patient.
13
| 09.11.19 - 01:52

Fluoroscopy

1.7 Case 7: Digital Subtraction 1.7.3 Discussion

Angiography and Motion
Artifacts
1.7.1 Background
● A patient undergoes a preoperative embolization
of renal cell carcinoma metastasis in the left hand.
● Digital subtraction angiography (DSA) is used to
demonstrate reduced blood flow to the tumor
following embolization.
1.7.2 Findings
● During the DSA run, the patient repositions her
hand, leading to significant artifacts.
● DSA was repeated to obtain adequate image
quality.
DSA is an imaging technique used to visualize
blood vessels while minimizing the appearance
of bone and soft tissues. It is performed by
acquiring a “mask” image of the anatomy of
interest before the administration of a contrast
medium (typically iodine-based), followed by a
series of postcontrast images. The mask image is
subtracted from each image in the postcontrast
series, reducing the visibility of overlapping
tissues. When properly utilized, the resulting
images provide a map of the vasculature of
interest and have several uses in identifying
stenosis, embolism, and aneurysm.
Proper application of this technique requires
that the anatomy being imaged is stationary in
the imaging field of view. ▶ Fig. 1.6a shows the
patient’s hand at the beginning of the DSA run
before contrast has been administered. The hand

Fig. 1.6 Individual frames from the

beginning (a), middle (b), and end
(c) of a DSA run where patient motion
caused significant artifacts are
shown. (d) The ability to visualize
the vasculature of the hand was
compromised leading to a repeated
acquisition.

14
| 09.11.19 - 01:52
1.7 Case 7: Digital Subtraction Angiography and Motion Artifacts
is well aligned with the previously acquired mask
image. Consequently, the subtraction image
masks the appearance of soft tissue and bone.
▶ Fig. 1.6b shows an image near the beginning of
contrast administration. The radial artery is
easily visualized as a result of contrast injection.
However, as contrast is administered, the
patient’s hand begins to move. This is most
readily observed at the fingers where the
misalignment between the current acquisition
and the mask image is greatest. Finally, ▶ Fig. 1.6c
shows an image that occurs near the end of the
DSA run where the hand is now substantially
mispositioned relative to the mask image. At this
point in the run, the vasculature of the hand
should be clearly visible but artifacts obscure the
anatomy of interest.
The initial position of the hand during the
acquisition of the mask image can be readily iden-
tified in ▶ Fig. 1.6c by the lighter, almost white
fingers and outer portion of the hand. Since the
position of the hand has changed, the mask image
is being subtracted from an area that no longer
contains the anatomy. Consequently, this portion
of the image is even brighter than the background.
As a corollary, a darker image of the thumb and
fingers is apparent, indicating the final position of
the hand.
1.7.4 Resolution
The example in ▶ Fig. 1.6 highlights the effects of
motion in an exam of the hand. Extremities may
be more prone to these artifacts since there is
higher possibility of movement. However, any
changes in patient position between acquisition of
the mask image and the subsequent images will
result in these effects. This can be caused by the
operator moving the patient, table, or X-ray tube,
or by involuntary patient motion (cardiac and res-
piratory). It should be noted that, in this case, the
DSA run was repeated with reduced patient
motion and image quality was greatly improved
(▶ Fig. 1.6d).
While extremities are not typically exposed to
large amounts of radiation during fluoroscopy
exams, procedures that are complex or require
imaging through larger portions of the body
require higher doses and have the potential to
result in radiation injuries. These risks are more
likely during DSA acquisitions compared to
conventional fluoroscopy since the dose rate
during DSA is substantially higher (even 20 times
higher). Good practice is to limit the use of DSA
to portions of the exam where it provides a clear
clinical benefit over fluoroscopy and avoid
repeating DSA acquisitions when possible.
15
| 09.11.19 - 01:52
Fluoroscopy
1.8 Case 8: Fluoroscopy Modes
and Dose
1.8.1 Background
● A patient with unstable angina received a cardiac
catheterization procedure under fluoroscopic
guidance.
● Standard fluoroscopy, digital spot images, and
cine mode fluoroscopy were used.
1.8.2 Findings
The reference AKR during cine angiography was
28.5 times higher than during standard fluoroscopy.
1.8.3 Discussion
Modern fluoroscopic systems have multiple oper-
ating modes, including standard fluoroscopy, high-
dose-rate (HDR) fluoroscopy, digital spot imaging,
and cine (sometimes referred to as cinefluorog-
raphy). The selection of an imaging mode during a
procedure requires consideration of the image
quality needs of the individual case as well as the
relative radiation exposure rate for each mode.
For all fluoroscopy systems, the maximum dose
rate that can be physically produced by the system
is limited by regulation for standard fluoroscopy
and HDR modes.11 The location along the X-ray
beam at which this is measured varies for different
types of fluoroscopy systems. (For under-table
systems, the maximum dose rate is measured at
1 cm above the patient table, while for C-arm type
systems used in interventional radiology and
cardiac catheterization labs, the maximum dose
rate is measured at 30 cm from the image
receptor.) When used in standard fluoroscopy
mode, the maximum allowable dose rate is 88
mGy/minute.11 In HDR mode the limit doubles to
176 mGy/minute. While standard fluoroscopy
produces adequate image quality for many
patients, HDR may be required to maintain image
quality in larger, more attenuating patients.
In digital spot mode, the fluoroscopy system
functions as a general radiography device
producing a single image. The technique (kV, mAs)
used to produce a digital spot image is generally
16
higher than that used for an individual fluoroscopy
frame producing a higher-quality image but with
an associated increase in dose. While the dose per
image is substantially higher in digital spot mode,
the total exposure to the patient may be less than
an individual fluoroscopy run which may contain
dozens of images.
Finally, cine mode creates an “X-ray movie”
similar to fluoroscopy but the image quality of
each individual frame is roughly equivalent to that
of a digital spot image. This improved image
quality for the entire cine run comes at the
expense of greatly increased dose rates. It should
be noted that the dose rate in cine mode is not
limited by regulation. This case illustrates the
difference in dose between standard fluoroscopy
and cine mode fluoroscopy.
▶ Fig. 1.7 shows two single-frame images of the
same anatomic location. Collimation, X-ray tube
and detector positions, and frame rate remained
the same for both images. ▶ Fig. 1.7a was acquired
as part of a fluoroscopy run, while ▶ Fig. 1.7b was
part of cine run. The cine image (▶ Fig. 1.7b)
clearly provides better image contrast as well
as improved detail of high spatial frequency struc-
tures (edges). During the fluoroscopy run, the
reference AKR was 20 mGy/minute. In contrast,
the reference AKR during the cine run was 570
mGy/minute (more than 28 times the AKR during
standard fluoroscopy).
1.8.4 Resolution
It is important for fluoroscopy operators to care-
fully consider whether the imaging task requires
the quality provided by cine and HDR fluoroscopy
modes. Commonly, positioning of the imaging field
of view relative to the patient can be performed
using standard fluoroscopy with higher-dose-rate
modes reserved for the anatomy of interest during
select portions of the procedures. Positioning can
also be aided by use of virtual collimators and
positioning tools.
Similarly, digital spot imaging yields a higher
dose per image than fluoroscopy. All modern fluo-
roscopy devices are equipped with a “last image
hold” function that presents the final image from
any fluoroscopy run to the operator, providing a
lower quality (and lower dose) static image. Digital
| 09.11.19 - 01:52

1.8 Case 8: Fluoroscopy Modes and Dose

Fig. 1.7 (a) A single frame from a fluoroscopy run performed during a cardiac catheterization procedure. (b) A cine
image from the same procedure using the same positioning and collimation. While image quality is improved on the
cine image, it comes at the expense of a substantial increase in dose rate. This image is similar to what would be seen
from a digital spot.

spot images may then be reserved for situations
where the improved image quality is necessary. As
is the case with other dose reduction practices,
imaging modes that result in lower patient dose
also reduce staff dose.
It should be noted that while the Food and
Drug Administration (FDA) places limits on the
maximum AKR that a system can produce while
operating in fluoroscopy mode,11 no such limit
exists when the system is used in other modes
(cine, DSA, or single-shot acquisitions). While the
AKR in standard fluoroscopy mode was well
below the regulated maximum of 88 mGy/minute
(176 mGy/minute when operated in high-dose-
rate fluoroscopy mode), the dose rate in cine
mode was considerably higher. Interestingly, the
reference point at which the maximum AK
is regulated by the FDA is often different from the
location of where the displayed reference AK
is calculated. Consequently, the maximum dis-
played AKR may be higher than the regulated
maximum AKR.

17
| 09.11.19 - 01:52
Fluoroscopy
1.9 Case 9: Equalization Filters
1.9.1 Background
● A patient with a history of primary sclerosing
cholangitis, coagulopathy, and chronic
portal vein thrombosis undergoes an
intrahepatic portosystemic shunt creation
procedure.
● Much of the procedure requires imaging over
an area that includes the lungs, liver, and bony
structures of the vertebral bodies.
1.9.2 Findings
● The lung appears hyperintense, obscuring
any anatomical contrast in the lungs
(▶ Fig. 1.8).
Fig. 1.8 The lung in this image appears extremely
bright. The image also contains space to the patient’s
left which also appears hyperintense. The image
contrast in other regions of the image is poor.
18
● The left portion of the image contains “open
space,” where there is no anatomy to attenuate
the X-ray beam.
● Contrast is poor inferior to the lungs.
1.9.3 Discussion
Section 1.4, Case 3: Collimation, discussed the use of
collimators to limit the X-ray field to the anatomy of
clinical interest. These collimators are made of lead
and are thick enough to fully attenuate the X-ray
beam. In addition to collimators, many fluoroscopy
systems have equalization filters. These filters are
sometimes known as wedges or contour filters. As
the name implies, wedge filters have varying thick-
ness and are typically made of lead-impregnated
acrylic sheets. They are less absorptive than collima-
tors and provide more attenuation through the
“heel” portion of the wedge than through the “toe.”
The filters can be moved in and out of the imaging
field view over structures that are less attenuating.
To understand the effect that equalization
filters have on image quality, one must consider
how digital images are displayed. The ultimate
purpose of these physical filters is to equalize the
signal intensity across the image. Since a finite
number of grayscale levels are available to
display a digital image, having a wide range of
signal intensities across the field of view results
in having fewer grayscale values available to
display less-attenuating anatomy, thus degrading
image contrast in these areas. Proper use of
wedge filters can help to maintain image contrast
for exams where the attenuation properties of
objects in the imaging field of view vary
substantially.
The case presented here highlights the effect of
not using equalization filters. The lungs
are less attenuating relative to other anatomic
structures in the image, that is, the diaphragm
and vertebral bodies. Without equalization
filters, the lungs appear extremely bright and
washed out.
1.9.4 Resolution
Proper use of wedge filters can improve image
quality, especially in cardiac and thoracic imaging,
| 09.11.19 - 01:52

1.9 Case 9: Equalization Filters

where an image often contains portions of the

heart, lungs, and bone, all of which have vastly
different X-ray attenuation properties. ▶ Fig. 1.9
provides an example of the benefits of using wedge
filters to equalize the signal intensity across the
image. The wedge filters are placed over portions of
the right and left lungs, improving image quality
across the entire image. An added benefit of using
equalization filters is a reduction in radiation dose
to areas covered by the filters.
Another way to achieve equalization of attenua-
tion properties across the image is to place a bolus
material next to the patient. In the case shown in
▶ Fig. 1.8, a bolus placed next to the patient’s chest
would decrease the brightness on the left side of the Fig. 1.9 An image showing wedge filters positioned
image. However, bolus materials increase scatter, over the left and right portions of the image.
By partially covering the radiotransparent lungs,
degrading image quality and increasing patient and
image quality is improved across the entire field
operator dose. For this reason, wedge filters are of view.
typically preferred to achieve image equalization.12

19
| 09.11.19 - 01:52
Fluoroscopy
1.10 Case 10: Cone Beam
Computed Tomography
1.10.1 Background
● A patient has a history of massive hemoptysis
and blood visualized in the right bronchus
during bronchoscopy.
● An embolization procedure is performed,
including a right bronchial artery angiogram.
● Cone beam computed tomography (CBCT) imaging
is used to provide axial and coronal images of the
liver and localize the tumor for therapy.
1.10.2 Findings
Streaking and motion artifacts are clearly visible
on both the axial (▶ Fig. 1.10a) and coronal
(▶ Fig. 1.10b) images reconstructed from the CBCT
acquisition.
1.10.3 Discussion
CBCT is used in the interventional environment to
produce cross-sectional and 3D reconstructions.
This intraprocedural imaging technique is used to
visualize anatomy and guide intervention. In some
procedures, such as transarterial chemoemboliza-
tion (TACE) for treatment of liver cancer, CBCT
Fig. 1.10 Axial (a) and coronal (b) images using cone beam computed tomography are shown. Both have noticeable
streaking and motion artifacts.
20
has been demonstrated to be superior to DSA for
visualization of liver tumors.13 CBCT images can
also be compared with images obtained prior to
the procedure using either MDCT or magnetic
resonance imaging.
Similar to standard multidetector CT (MDCT),
CBCT utilizes projections taken at multiple angles
around the anatomy of interest to reconstruct
axial, sagittal, and coronal slices. As with other
tomographic techniques, anatomic visualization is
improved by removing overlapping tissue. Since
the principles of CBCT are similar to those for
MDCT, CBCT is susceptible to similar image arti-
facts. However, the clinical implementation of
CBCT may result in unique manifestations of these
artifacts.
▶ Fig. 1.10 shows axial (▶ Fig. 1.10a) and coronal
(▶ Fig. 1.10b) images using CBCT during a liver
embolization procedure. In the axial image,
streaking artifacts result from interactions
between the primary X-ray beam and the metal
catheter placed as part of the procedure. Incident
X-rays are highly absorbed by the catheter,
resulting in fewer X-ray photons reaching the
detector. The star pattern is due to this effect
occurring in multiple X-ray projections. In addi-
tion, in the coronal image, a possible beam
hardening artifact is shown inferior to a region of
hyperintensity.
The images in ▶ Fig. 1.10 also suffer from blurri-
ness, especially visible in the lungs which is caused
| 09.11.19 - 01:52

1.10 Case 10: Cone Beam Computed Tomography

by breathing motion. While motion artifacts also
occur during MDCT, the acquisition times of CBCT
tend to be longer, leaving more time for motion to
occur. (On modern systems, CBCT acquisition times
are generally around 6–10 seconds. Rotation times
in MDCT are often below 1 second.)
1.10.4 Resolution
Adequate patient breath holds, starting a couple of
seconds prior to initiation of imaging, are essential
during CBCT. ▶ Fig. 1.11 shows axial and coronal
images obtained using a patient breath hold.
Compared with the images shown in ▶ Fig. 1.10,
motion artifacts are markedly reduced. As with
MDCT, the ability to achieve an adequate breath
hold may be limited by the patient and procedure
being performed. An additional consideration is
that patients are more likely to be under some
level of sedation during CBCT, so the amount of
sedation should be considered in evaluating breath
hold options and limitations.
Compared with MDCT, CBCT suffers from low
contrast-to-noise ratio. However, this limitation
is not generally clinically significant because the
use of iodinated contrast greatly increases the
visibility of the relevant anatomy. CBCT tends to
have better inherent spatial resolution, since the
pixel pitch of the flat panel detectors on inter-
ventional angiography systems is smaller than
the detectors used in MDCT. (A modern flat panel
detector has a pixel pitch of approximately
0.15 mm; a MDCT detector is often around
0.6 mm.) Finally, studies have found that CBCT
results in lower patient skin dose compared with
DSA.13 CBCT also often results in shorter overall
imaging times, further reducing radiation dose to
both patients and staff.

Fig. 1.11 Axial (a) and coronal (b) images using cone beam computed tomography with breath hold are shown.
These images have fewer motion artifacts than those shown in ▶ Fig. 1.10.

21
| 09.11.19 - 01:52
Fluoroscopy
1.11 Review Questions
1.11.1 Case 1: SID, ABC, and
Radiation Output
1. During a fluoroscopy-guided interventional
procedure, the operator increases the SID from
90 to 120 cm without changing SSD. What effect
will this change have on entrance skin dose?
a) Entrance skin dose will increase by greater
than 50%.
b) Entrance skin dose will increase by less than
50%.
c) Entrance skin dose will be unaffected.
d) Entrance skin dose will be reduced by less
than 50%.
e) Entrance skin dose will be reduced by
greater than 50%.
2. When moving from imaging a patient’s torso
to imaging the abdomen, the ABC algorithm
used by the fluoroscopy system will likely
compensate by doing which of the following?
a) Lowering kV
b) Increasing mA
c) Adding filtration
d) Decreasing magnification
1.11.2 Case 2: Reference Air
Kerma and Skin Dose
3. During a pain management procedure using
a C-arm with a fixed SID, the operator raises
the patient table such that the SSD changes
from 50 to 75 cm. What effect will this change
have on entrance skin dose?
a) Entrance skin dose will increase by greater
than 50%.
b) Entrance skin dose will increase by less than
50%.
c) Entrance skin dose will be unaffected.
d) Entrance skin dose will be reduced by less
than 50%.
e) Entrance skin dose will be reduced by
greater than 50%.
22
4. If the patient table is raised from a height
of 50 cm above the X-ray source to 60 cm
above the X-ray source, and all other imaging
parameters remain the same, how much will
the displayed AK change?
a) The displayed AK will be 1.4 times the
AK displayed at 50 cm.
b) The displayed AK will be 1.2 times the AK
displayed at 50 cm.
c) The displayed AK will remain the same.
d) The displayed AK will be 0.8 times the
AK displayed at 50 cm.
e) The displayed AK will be 0.7 times the AK
displayed at 50 cm.
1.11.3 Case 3: Collimation
5. For a procedure performed using a single field
(i.e., without any tube rotation), collimating the
imaging field of view from 600 to 400 cm2 will
have what effect on PSD?
a) PSD will decrease approximately by 33%.
b) PSD will not change.
c) PSD will increase approximately by 33%.
6. For a procedure performed using a single field,
collimating the imaging field of view from 600
to 400 cm2 will have what effect on scatter dose
to the operator?
a) Operator dose will decrease by 33%
b) Operator dose will not change.
c) Operator dose will increase by 33%.
1.11.4 Case 4: Anti-scatter Grids
7. For what clinical exam would it be appropriate
to remove the anti-scatter grid?
a) Sacroiliac joint injection
b) Endoscopic retrograde
cholangiopancreatogram
c) Inferior vena cava filter placement
d) Upper extremity arteriovenous fistulagram
| 09.11.19 - 01:52
8. What is the primary consequence of removing
the anti-scatter grid for an adult abdomen
protocol?
a) Decreased image quality
b) Increased patient skin dose
c) Decreased visualization of iodine contrast
d) Increased DAP
1.11.5 Case 5: Patient Shielding
9. What is one of the most effective ways to
decrease radiation exposure to a fluoroscopy
operator?
a) Wear appropriate lead garments
(e.g., apron, glasses).
b) Position a lead drape on the patient.
c) Increase the SID.
10. When performing a cardiac fluoroscopy-guided
procedure on a pregnant patient, what is the
primary source of radiation exposure to the
fetus?
a) Primary X-ray beam
b) Internal scatter
c) Backscatter
1.11.6 Case 6: CT Fluoroscopy
11. What is the primary reason the DLP for a
CT-guided fluoroscopy-guided procedure is
typically higher than for a diagnostic CT scan?
a) The total CTDIvol is typically higher for
a CT fluoroscopy-guided procedure.
b) The scan length is typically higher for
a CT fluoroscopy-guided procedure.
c) A higher kV is typically used for a CT
fluoroscopy-guided procedure.
d) A higher effective mAs is typically used for
a CT fluoroscopy-guided procedure.
12. Radiation-induced skin injuries may occur at
PSDs as low as 2000 mGy. For an average-sized
adult patient, what CTDIvol would be associated
with a PSD of 2000 mGy?
a) 1300 mGy
b) 2000 mGy
1.11 Review Questions
c) 3000 mGy
d) 6000 mGy
1.11.7 Case 7: Digital Subtraction
Angiography and Motion Artifacts
13. In DSA, why are blood vessels readily visible
while the appearance of other tissues is
minimized?
a) The kV of the mask and second acquisitions
are adjusted to specifically remove certain
tissues types following subtraction.
b) The addition of contrast agents to the blood
vessels between the mask and the second
acquisition highlights the vessels.
c) The attenuation properties of blood vessels
are significantly different than those of other
soft tissue, making them more visible.
14. During DSA, an X-ray attenuating object is
removed from the imaging field of view after
the mask image is acquired but before the
second acquisition. What effect will it have on
the images resulting from subtraction?
a) The object will leave a dark shadow relative
to background.
b) The object will have no effect on the
subtraction image since it is removed before
the second acquisition.
c) The object will leave a light shadow relative
to background.
1.11.8 Case 8: Fluoroscopy Modes
and Dose
15. The patient dose from a digital spot image is
roughly equivalent to which of the following
images?
a) The dose from a single fluoroscopy frame
b) The dose from an HDR fluoroscopy frame
c) The dose from a single cine frame
16. For what imaging mode is the maximum AKR
limited to 176 mGy/minute?
a) Standard fluoroscopy
b) HDR fluoroscopy
c) Cine mode fluoroscopy
23
| 09.11.19 - 01:52
Fluoroscopy
1.11.9 Case 9: Equalization Filters
17. During a cardiac imaging procedure, the lungs
appear so bright on the images that the cardiac
structures are poorly visualized. What is the
most appropriate method for improving image
quality?
a) Positioning a bolus material under the
patient
b) Using equalization filters over the lungs
c) Decreasing the amount of copper filtration
d) Increasing the collimated field of view
18. What is the primary benefit of using equaliza-
tion filters?
a) Improved image contrast
b) Increased patient dose
c) Improved spatial resolution
d) Increased temporal resolution
1.11.10 Case 10: Cone Beam
Computed Tomography
19. What is the primary advantage of using CBCT
instead of DSA to view liver tumors during
interventional procedures?
a) Decreased patient skin dose
b) Decreased motion artifacts
c) Improved spatial resolution
d) Improved visualization of lesions
20. What is the primary limitation of CBCT
compared with MDCT?
a) Poor spatial resolution
b) Increased patient dose
c) Limited imaging field of view
d) Introduction of streaking artifacts
24
References
[1] AAPM Report 125. Functionality and operation of fluoro-
scopic automatic brightness control/automatic dose rate
control logic in modern cardiovascular and interventional
angiography systems (2012)
[2] Marsh RM. What considerations should be made when
performing fluoroscopy-guided interventions on pregnant
patients? AJR Am J Roentgenol. 2017; 209(3):W195–W196
[3] International Commission on Radiation Protection. Biological
effects after prenatal irradiation (embryo and fetus). ICRP
Publication 90. Ann ICRP. 2003; 33:1–206
[4] Dauer LT, Thornton RH, Miller DL, et al. Society of Interven-
tional Radiology Safety and Health Committee, Cardiovascu-
lar and Interventional Radiology Society of Europe Standards
of Practice Committee. Radiation management for interven-
tions using fluoroscopic or computed tomographic guidance
during pregnancy: a joint guideline of the Society of Inter-
ventional Radiology and the Cardiovascular and Interven-
tional Radiological Society of Europe with Endorsement by
the Canadian Interventional Radiology Association. J Vasc
Interv Radiol. 2012; 23(1):19–32
[5] Smith JR, Marsh RM, Silosky MS. Is lead shielding of patients
necessary during fluoroscopic procedures? A study based on
kyphoplasty. Skeletal Radiol. 2018; 47(1):37–43
[6] Petranovic M, Gilman MD, Muniappan A, et al. Diagnostic
yield of CT-guided percutaneous transthoracic needle biopsy
for diagnosis of anterior mediastinal masses. AJR Am J Roent-
genol. 2015; 205(4):774–779
[7] Leng S, Christner JA, Carlson SK, et al. Radiation dose levels
for interventional CT procedures. AJR Am J Roentgenol. 2011;
197(1):W97–103
[8] Kanal KM, Butler PF, Sengupta D, Bhargavan-Chatfield M,
Coombs LP, Morin RL. U.S. diagnostic reference levels and
achievable doses for 10 adult CT examinations. Radiology.
2017; 284(1):120–133
[9] Balter S, Hopewell JW, Miller DL, Wagner LK, Zelefsky MJ.
Fluoroscopically guided interventional procedures: a review
of radiation effects on patients’ skin and hair. Radiology.
2010; 254(2):326–341
[10] AAPM Report 96. The measurement, reporting, and manage-
ment of radiation dose in CT (2008)
[11] Performance standards for ionizing radiation emitting prod-
ucts, 21 C.F.R. §1020.32 (2018)
[12] Schueler BA. The AAPM/RSNA physics tutorial for residents:
general overview of fluoroscopic imaging. Radiographics.
2000; 20(4):1115–1126
[13] Tacher V, Radaelli A, Lin M, Geschwind JF. How I do it: cone-
beam CT during transarterial chemoembolization for liver
cancer. Radiology. 2015; 274(2):320–334
| 09.11.19 - 01:52
2 Mammography
Ingrid S. Reiser
Introduction
Mammography uses X-ray projection imaging to
acquire images of the breast. Mammography has
been the gold standard for breast cancer scree-
ning since the 1990s. In addition to breast cancer
screening, mammography is used for diagnostic
breast imaging. Conventional mammography
produces 2D projection images of the breast. The
complex anatomical background of structures can
hinder tumor detectability. With the advent of
digital full-field detectors for mammography,
digital breast tomosynthesis (DBT) became
feasible. DBT produces quasi-3D images of the
breast that provide some depth resolution to
overcome limitations of geometric superposi-
tioning in conventional 2D projection images.
DBT has helped reduce callback rates from scree-
ning mammography and potentially helps
increase cancer detection rates.
Mammography utilizes ionizing radiation to
detect breast cancers. Since ionizing radiation also
induces cancer (albeit at a much lower rate), it is
important to limit and monitor the average dose
to glandular breast tissues during mammography.
Common Image Quality
Problems
Mammography is a unique imaging modality in
the sense that it is dedicated to and optimized for
a single anatomy. The energy of the X-ray beam,
X-ray absorption, and resolution of the detector
are optimized for the detection of low-contrast
tumors and microcalcifications in the breast. X-ray
energies impact radiographic contrast and patient
dose, and are optimized to produce images with
the best tumor signal-to-noise ratio at the lowest
dose. Over- or underexposure of the mammogram
due to incorrect X-ray technique is less common in
digital detectors than film because of their wide
dynamic range. However, there are a number of
detector artifacts in digital mammography that
can potentially mimic or obscure suspicious find-
ings that must be recognized. In DBT, a quasi-3D
image volume is synthesized from a series of low-
dose projection images acquired with the X-ray
tube moving across an arc of 15 to 50 degrees
(depending on vendor implementation). The
limited angle acquisition provides some depth
resolution, but also causes artifacts in the tomo-
synthesis image volume.
As in radiography, the most common problem
affecting image quality is patient positioning and
patient motion. Further, image quality can be
affected by artifacts that mimic or obscure relevant
anatomy. Speck-like artifacts can potentially
mimic microcalcifications. There are multiple
causes for these artifacts as described below. Arti-
facts caused by the image receptor include row
dropout and electromagnetic interference (EMI).
Mammography utilizes a grid in contact mode.
Artifacts due to grid positioning, such as grid
cutoff, do not occur because the grid and X-ray
source are in a fixed geometry.
25
| 09.11.19 - 01:52
Mammography
2.1 Case 1: Magnification
Imaging
2.1.1 Background
Magnification views are acquired when visualiza-
tion of fine detail is required. Magnification views
are often used in the diagnosis of microcalcifica-
tions. ▶ Fig. 2.1 demonstrates this effect. A micro-
calcification cluster is imaged in magnification
mode (▶ Fig. 2.1a), which is shown enlarged in
▶ Fig. 2.1b. ▶ Fig. 2.1c shows the microcalcification
cluster imaged in contact mode. The resolution in
▶ Fig. 2.1c is markedly lower and as a result less
detail is perceived.
2.1.2 Findings
The images shown in ▶ Fig. 2.1a, b were acquired
with the breast placed on the magnification
stand at a height of 31 cm, resulting in a magnifi-
cation factor of 1.8 for a source-to-detector
distance of 70 cm. An increase in resolution is
Fig. 2.1 Magnification views improve visualization of fine detail. Microcalcification cluster magnification imaging
(a, b) and in contact imaging (c). The magnification factor in this example was 1.8. Geometry of magnification imaging:
object magnification (d).
26
gained from the magnification setup that
includes use of a smaller focal spot compared to
contact mode to minimize focal spot blurring.1
2.1.3 Discussion
In magnification mode, the breast is positioned
on a magnification stand and is thus located
closer to the X-ray source (▶ Fig. 2.1d). As a
result, structures within the breast are magnified
by a factor M, depending on the distance between
the X-ray focal spot and the breast on the magni-
fication stand (dmag), and the distance between
the X-ray focal spot and the detector (ddet). The
magnification factor M is given by the ratio of
these distances as M = ddet/dmag.
2.1.4 Resolution
The height of the magnification stand deter-
mines the magnification factor. Most mammog-
raphy systems allow for several heights of the
magnification stand.
| 09.11.19 - 01:52

2.2 Case 2: Focal Spot Size Selection in Magnification Views

2.2 Case 2: Focal Spot Size
Selection in Magnification
Views
2.2.1 Background
● Magnification views can help visualize small
structures, such as microcalcification clusters,
with greater detail. In contact imaging, when the
breast is placed directly on the breast support, the
system uses a large focal spot (typically 0.3 mm).
● When the breast is positioned on the magnifi-
cation stand and magnification mode is used,
the system switches to a small focal spot
(typically 0.1 mm).
2.2.2 Findings
▶ Fig. 2.2a, b shows a comparison of a test phantom
imaged with a large (0.3 mm) and a small (0.1 mm)
focal spot. When the phantom is placed on the
breast support (▶ Fig. 2.2a), the images acquired
with a large or small focal spot exhibit similar
sharpness. When the test phantom is placed on the
magnification stand, the image acquired with the
large focal spot is less sharp due to focal spot blur.
2.2.3 Discussion
The focal spot of an X-ray tube is not a single
point, but a small area from where X-rays are
emitted. In mammography, the size of the focal
spots are 0.3 mm (large) and 0.1 mm (small). The
finite size of the focal spot results in blurring of
the image.1 This is demonstrated in ▶ Fig. 2.2a, b
and depends on the distance between the focal
spot, the breast, and the detector. When the
object (i.e., breast) is located in close contact with
the detector and the focal spot is far away, focal
spot blur is negligible for both the large and the
small focal spot. Therefore, the images shown in
▶ Fig. 2.2a are equally sharp. In magnification

Fig. 2.2 (a) Phantom images acquired in contact mode with different focal spot sizes (FS). (b) Phantom images
acquired in magnification mode with different focal spot sizes (FS). (c) Focal spot blur in contact mode and
magnification mode.

27
| 09.11.19 - 01:52

Mammography

mode, the breast is half-way between X-ray focal blur. Focal spot blur can be computed from the
spot and the detector. In this configuration, the imaging distances as:
size of the projected focal spot is equal to the

ddet  dmag
actual focal spot size and blurs the image accord- wblur ¼ w 
dmag
ingly. This is shown in ▶ Fig. 2.2b.
The differences in blur become visible when where w is the size of the focal spot, and wblur is
focal spot blur is the dominant factor that limits the width of focal spot blur.
image resolution. In contact imaging, shown in
▶ Fig. 2.2a, image sharpness is limited by
detector resolution, and therefore no difference 2.2.4 Resolution
in image sharpness is observed when the focal
spot sizes are changed. The small focal spot needs to be selected when
▶ Fig. 2.2c shows how the distance of the object performing imaging in magnification mode to
from the X-ray focal spot, dmag, affects focal spot optimize resolution of the system.

28
| 09.11.19 - 01:52
2.3 Case 3: X-ray Acquisition Technique Factors in Mammography
2.3 Case 3: X-ray Acquisition
Technique Factors in
Mammography
2.3.1 Background
In mammography, technique factors, such as mAs,
kVp, filtration, are chosen depending on
compressed breast thickness and composition
(fatty/heterogeneous/dense). Higher kV is chosen
for larger breast thicknesses.
2.3.2 Findings
● Mammograms of two different patients are
shown in ▶ Fig. 2.3a, b. The compressed breast
thicknesses are 35 and 82 mm. The patients were
imaged on the same mammography system
using automated exposure control.
● X-ray technique factors (mAs, kV, filter) for the
acquisition of these two images are shown in
the figure, along with half-value layer (HVL)
and average glandular dose (AGD).
● The system selected higher mAs, kV, and a
different filter to image the 82-mm breast,
resulting in a greater HVL and AGD.
● The ACR digital mammography accreditation
phantom was imaged at equal average glandular
dose but different kV and target/filtration
settings, and image quality was compared
(▶ Fig. 2.3c, d).
2.3.3 Discussion
As the X-ray tube potential (i.e., kV) increases,
subject contrast is reduced. This is demonstrated
in ▶ Fig. 2.3c, where the contrast of all masses is
greater for the 28 kV W/Rh image, compared to
37 kV W/Ag. As the X-ray tube potential (i.e., kV)
increases, contrast-to-noise ratio reduces, as
shown in ▶ Fig. 2.3d. The display window width
is set to achieve equal displayed contrast of the
largest mass, compared to the background. The
37-kV image is noisier, and it is more difficult to
see the smaller masses and fibers. Technique
factors are set so as to optimize the contrast-
to-noise ratio per average glandular dose. For
Fig. 2.3 (a, b) Mammographic
technique for different breast sizes.
(c, d) Signal insert of the full-field
digital mammography accreditation
phantom imaged at equal average
glandular dose (AGD), but different
filtration and kV settings. In (c), the
display window width is equal for both
images. In (d), the display window is
different for both kV settings and was
chosen to produce equal contrast of
the largest mass in both images.
29
| 09.11.19 - 01:52
Mammography
the phantom images, the optimum X-ray tech-
nique is 28 kV compared to 37 kV, because it
gives better image quality in terms of higher
contrast and contrast-to-noise ratio. When
imaging a thicker breast, generally a higher kV is
used to achieve a more penetrating beam. This is
manifested in the increased HVL of the X-ray
beam. Improved image quality could also be
achieved by increasing mAs to maintain subject
contrast, but prolonged exposure times can lead
to artifacts from patient motion (see Case 9, Pa-
tient Motion Causing Blurred Parenchymal
30
Structure in a Mammogram), and it can also
potentially cause X-ray tube heat overloading. In
addition, increasing mAs increases dose. Average
glandular dose is linearly proportional to mAs, i.
e., doubling mAs produces twice the AGD.
2.3.4 Resolution
kV should be increased with breast thickness and
breast density to produce a more penetrating
X-ray beam that optimizes subject contrast.
| 09.11.19 - 01:52
2.4 Case 4: Digital Breast
Tomosynthesis: Artifacts due
to High-Contrast Objects
2.4.1 Background
Tomosynthesis imaging is a quasi-3D imaging
modality of the breast. The X-ray source travels
along an arc of 15 to 50 degrees, depending
on vendor implementation, and a series of low-
dose projections are acquired. The tomosyn-
thesis images are reconstructed from these
projections.
2.4.2 Findings
● ▶ Fig. 2.4a shows a large calcification as it
appears in the focus plane, i.e., in the tomosyn-
thesis image at the actual depth of the object.
The insets show slices displaying the calcifica-
tion at different depths. The calcification is
clearly seen at other depths, but it becomes more
distorted as the tomosynthesis image is further
away from the in-focus depth.
● ▶ Fig. 2.4b shows a perpendicular slice through
the breast volume. The calcification is clearly
Fig. 2.4 (a) Artifact due to a high-contrast calcification in tomosynthesis volume. Regions centered on the calcification
are shown at different depths. Repeated ghosts of the calcification can be observed at depths far from its in-focus plane.
(b) Perpendicular slice through the tomosynthesis volume at the level of the high-contrast calcification. The X-ray tube
moves along the X-direction during the tomosynthesis scan.
2.4 Case 4
seen as it fans out as the distance to the focus
slice increases. The angle of the fan is that of the
tomosynthesis scan.
2.4.3 Discussion
In tomosynthesis, depth resolution is achieved by
blurring structures below and above the in-focus
plane, and enhancing structures that are truely
located at that depth. High-contrast structures
often persist throughout all tomosynthesis images
and present themselves as repeat “copies” of the
object, as is the case for the microcalcification
shown in ▶ Fig. 2.4a. Sometimes, a calcified vessel
appears as ripples in slices above or below the slice
of focus. Other high-contrast objects that can
produce such artifacts include biopsy markers.
2.4.4 Resolution
The limited angle scan geometry of tomosynthesis
causes artifacts from high-contrast objects, such as
large calcifications. Depending on the image
reconstruction and image processing used by the
equipment manufacturer, conspicuity of these
artifacts can vary. It is important to understand
the origin of these artifacts.
31
| 09.11.19 - 01:52

Mammography

2.5 Case 5: Effect of Image
Post-Processing on the
Appearance of a Mammogram
2.5.1 Background
● Digital mammography uses image processing
to enhance contrast.2,3
● Processed mammograms are overall more
uniform in comparison to a screen-film
mammogram, which virtually eliminate the
need to adjust display window width and
level.
● Mammograms obtained from different vendor
systems can have a significantly different “look”
due to proprietary choices in image processing
and presentation.
2.5.2 Findings
● ▶ Fig. 2.5a–d shows images from a patient
acquired on different vendor systems. ▶ Fig. 2.5a
shows a screen-film mammogram. The intensity
in the image falls off toward the skin line because
breast thickness decreases.
● ▶ Fig. 2.5b shows a mammogram of the same
patient but acquired on full-field digital
mammography (FFDM). The breast is visual-
ized up to the skin line, which was achieved
through a uniformity correction. As a result,
structures close to the skin line are still clearly
visible.
● ▶ Fig. 2.5c shows a mammogram acquired
on a FFDM unit of a different vendor. As in
▶ Fig. 2.5b, the breast is visualized up to the
skin line. However, the image is processed
to show greater contrast compared to that
shown in ▶ Fig. 2.5b. ▶ Fig. 2.5d shows a
synthetic 2D image generated from tomosyn-
thesis volume images for the same patient as
in ▶ Fig. 2.5c. The synthetic mammogram is
similar but not equal to a conventional 2D
mammogram. In the example shown, the calci-
fication exhibits dark overshoots above and
below it, which is an artifact of tomosynthesis

Fig. 2.5 (a–d) Image processing affects the appearance of mammograms. A patient was imaged with different systems:
screen-film mammography (a), full-field digital mammography (FFDM) vendor A (b), FFDM vendor B (c), 2D
mammogram synthesized from tomosynthesis (d). Note the difference in the calcification (arrows) shown in the inset in
(c) and (d). (Continued)

32
| 09.11.19 - 01:52

2.5 Case 5: Effect of Image Post-Processing on the Appearance of a Mammogram

Fig. 2.5 (Continued) (e) Fourier-domain

filters can enhance large or small detail
in an image.

Fig. 2.5 (Continued) (f–h) Original mammogram (f) processed with different unsharp masking parameters (g, h). The
image uniformity is improved, which allows display of the image at a higher display contrast in (h), while maintaining
visibility of the skin line.

33
| 09.11.19 - 01:52
Mammography
that is not seen in the conventional 2D
mammogram (▶ Fig. 2.5c).
2.5.3 Discussion
In digital mammography as in digital radiog-
raphy, the image consists of an array of numeric
values. Image processing is achieved through
mathematical operations on these numbers (i.e.,
pixel values). Two basic image processing opera-
tions, by use of the Fourier transform and by
histogram processing, are demonstrated below.
The Fourier transform is a basic element of image
processing that takes advantage of the spatial fre-
quency. Low spatial frequencies represent large
structures and image contrast, while high spatial
frequencies represent image detail such as fine
structures and edges, and image noise. A low-
pass filter selects large structures in an image,
while a high-pass filter selects image detail and
34
enhances edges. The effect of these filters is
demonstrated in ▶ Fig. 2.5e.
In unsharp masking, a low-pass filtered copy
of the original image is subtracted from the image,
producing a sharper and more uniform image
▶ Fig. 2.5f-h. Digital mammograms are processed
to achieve greater uniformity, eliminating the
intensity drop-off near the skin line that
is observed in screen-film mammograms. In addi-
tion, images might be processed to produce a
greater displayed contrast.2
2.5.4 Resolution
Image processing algorithms are proprietary and
can differ between mammography equipment
vendors. Differences in image appearance between
vendors are more likely caused by different image
post-processing algorithms than differences in
acquisition technique or dose.
| 09.11.19 - 01:52

2.6 Case 6: Artifact due to Detector Row Dropout

2.6 Case 6: Artifact due to
Detector Row Dropout
2.6.1 Background
Detector artifacts can be caused by individual pixel
dropout or by an entire detector row dropout.4,5
This case describes the appearance of detector row
dropout in conventional 2D mammography and 3D
tomosynthesis and associated synthetic 2D view.
2.6.2 Findings
● In this example, dropout is observed for two
detector rows. ▶ Fig. 2.6a shows the appearance
of the artifact in the conventional mammogram.
As expected, two lines are seen.
● ▶ Fig. 2.6b shows the appearance of the artifact
in 3D tomosynthesis. In this particular case,
row dropout occurred at two locations in one
single projection view. In the tomosynthesis
images, two lines can be observed and the loca-
tion of the lines changes in different depths. In
the simulated 2D image, which incorporates
information from the 3D volume, the row
dropout forms a band corresponding to the line
dropout visualized in the tomosynthesis
images at different depths.
2.6.3 Discussion
The appearance of the dropout artifact differs
in the conventional 2D mammogram and the
simulated 2D image. This underlines the fact that
the synthetic 2D image is created from the 3D
tomosynthesis volume. Depending on vendor
implementation, the synthetic 2D image may not
be intended to mimic a 2D mammogram, but
instead enhances potentially suspicious features
that are seen in the tomosynthesis image, such as
edges. In this case, edge post-processing likely
enhanced the line artifacts and caused the banding
in the synthetic 2D image.
2.6.4 Resolution
The dropout signal observed in these images
represents a failure in the readout hardware of the
digital detector and could not be remediated by
recalibrating the detector. In this case, the detector
panel needed to be replaced.

Fig. 2.6 (a) Line artifacts due to detector row dropouts in a conventional mammogram. In this case, two lines are seen.
(b) Line artifacts due to detector row dropout in tomosynthesis. In this example, row dropout occurred in a single
projection view. In the tomosynthesis images, two lines are seen. The lines sweep across the tomosynthesis images
while scrolling through different depths. In the corresponding synthetic 2D image, the lines appear as bands.

35
| 09.11.19 - 01:52

Mammography

2.7 Case 7: Microcalcification-like 2.7.3 Discussion

Appearance Caused by a Detector Small speck-like artifacts can be observed in
mammograms. Their cause can be multiple (also see
Artifact Case 8, Artifact due to Imperfection in Compression
2.7.1 Background Paddle). An artifact in the detector manifests itself in
the same location in different images. This artifact
● An unexpected microcalcification-like object can be distinguished from dust or scratches by
was observed during ACR phantom image wiping all surfaces and rotating the test phantom.
quality review. An artifact in the detector will always remain at the
● Prior patient images were reviewed. An same location. Multiple quality control tests might
reveal this artifact, including a weekly artifact test
artifact was found at the identical pixel
during which an image of a uniform phantom is
locations. inspected, as well as the phantom image quality test.

2.7.2 Findings 2.7.4 Resolution

● The defect is seen in prior patient images at the
same pixel location.
● In the clinical example shown below, the artifact
appears dark, while it appears white in the
phantom image.
Detector artifacts can be resolved by updating
the dead pixel map, which is used to mask out
bad pixels. If there are too many bad pixels
or the region of bad pixels is too large, the manu-
facturer might choose to replace the detector
(▶ Fig. 2.7).

Fig. 2.7 Dead pixel artifact in the detector in (a) patient image and (b) ACR mammography accreditation phantom
image. Artifacts in the detector occur at the same location in the patient image, and they might appear as dark or
bright, or mixed (see insets).

36
| 09.11.19 - 01:52

2.8 Case 8: Artifact due to Imperfection in Compression Paddle

2.8 Case 8: Artifact due to
Imperfection in Compression
Paddle
2.8.1 Background
● The radiologists noticed a small speck-like
artifact within the mammogram, with
microcalcification-like appearance, that
was seen in most images from a particular
unit.
● A compression paddle had a crack and was
replaced with a new paddle a few days prior.
2.8.2 Findings
● A small artifact is seen (▶ Fig. 2.8a). It is visible
in most but not all patient images, and at
different locations.
● A small speck (< 1 mm in diameter) can be seen
with use of the new compression paddle.
2.8.3 Discussion
The artifact is only seen in the patient images for
which this particular compression paddle was
used. The artifact does not appear in the same
location because its projection shifts across the
detector as the compression paddle height is
adjusted according to breast thickness
(▶ Fig. 2.8b).
2.8.4 Resolution
An imperfection on the compression paddle mani-
fests as a high-contrast microcalcification-like
appearance in patient images. If cleaning the
paddle does not remedy the artifact, the compres-
sion paddle should be replaced.

Fig. 2.8 (a) Location of the artifact in two mammograms with different compressed breast thicknesses. The red line
indicates the location of the same detector row in both images. (b) The projected location of the artifact changes with
compression paddle height.

37
| 09.11.19 - 01:52
Mammography
2.9 Case 9: Patient Motion
Causing Blurred Parenchymal
Structure in a Mammogram
2.9.1 Background
● This left-medial lateral oblique (LMLO)
mammogram is part of a screening exam
of an asymptomatic patient.
● Breast cancer screening exams are rated using
the BI-RADS scale. This exam resulted in a
BI-RADS score of 0 (recall) for technical reasons.
● The patient returned to the clinic and a technical
repeat of this view was performed.
2.9.2 Findings
● The parenchymal structure of the breast appears
blurred due to patient motion.
● Microcalcification clusters, which represent an
important early indicator of breast cancer, might
be missed due to insufficient spatial resolution
of this image.
● Exposure times in mammography are relatively
long. This image was acquired with a 1.5-second
X-ray exposure.
Fig. 2.9 (a) Left-medial lateral oblique mammogram with signs of patient motion. The image appears unsharp.
Fine detail of the parenchymal structure is blurred due to motion. (b) Repeated image with increased compression.
Fine linear details in the parenchymal structure are visible in this image.
38
● Original image: compressed breast thickness
was 72 cm, with 14 lb compression force. Repeat
image: compressed breast thickness was 60 cm,
with 23 lb compression force.
2.9.3 Discussion
Mammography requires high spatial resolution
in order to depict clusters of microcalcifications.
Individual microcalcifications of clinical
relevance have diameters ranging from 100 to
500 μm. In mammography, X-ray images are
acquired with the breast under compression
with typical compression forces between 25 and
45 lb. The technologist has to balance patient
comfort and image quality. Too much compres-
sion force leads to patient discomfort. Too little
compression force reduces breast immobiliza-
tion and can result in patient motion, as seen in
this case.
2.9.4 Resolution
A repeat mammogram was obtained (▶ Fig. 2.9b).
The technologist applied a higher compression
force, resulting in better breast immobilization
and the absence of motion artifacts.
| 09.11.19 - 01:53

2.10 Case 10: EMI Artifact due to LVAD Device

2.10 Case 10: EMI Artifact due 2.10.3 Discussion

to LVAD Device
2.10.1 Background
A patient received a routine mammogram. Due other
health reasons, the patient has a left-ventricular
assist device (LVAD).
2.10.2 Findings
The mammogram (▶ Fig. 2.10) exhibits high-
frequency striping near the chest wall edge.
Visually, this artifact has similarity with grid
artifacts that can be observed in radiography.
This EMI artifact occurs when electromagnetic
fields emitted by the LVAD motor interfere with
the detector readout electronics.
2.10.4 Resolution
The radiologist needs to be aware that the pres-
ence of the LVAD devices in patients can cause
such artifacts. Unless the mammography equip-
ment manufacturer has implemented solutions
into their imaging system to prevent such inter-
ference through shielding or image processing
designed to remove these structures, these
artifacts need to be tolerated.

Fig. 2.10 (a) Mammogram with the left-ventricular assist device (LVAD) device partially visible. When zoomed into
a region near the chest wall (b), horizontal lines can be seen that are caused by the electromagnetic interference
between the LVAD motor and the detector readout electronics.

39
| 09.11.19 - 01:53

Mammography

2.11 Review Questions 2.11.3 Case 3: X-ray Acquisition

2.11.1 Case 1: Magnification Technique Factors in
Imaging Mammography
5. What is the most likely range of X-ray tube
1. A small microcalcification (diameter = 100 μm)
potentials used in mammography?
is imaged in contact mode. What is the approx-
a) 15–25 kV
imate diameter of the microcalcification in the
b) 25–35 kV
image? Assume dmag = 67.5 cm and ddet = 70 cm.
c) 35–45 kV
a) 100 µm
d) 45–50 kV
b) 200 µm
c) 400 µm
6. Subject contrast in a mammogram is most
d) 500 µm
affected by which of these parameters?
2. A small microcalcification (diameter = 100 μm) is a) Milliampere-time product (mAs)
imaged in magnification mode. The detector b) X-ray tube kilovoltage (kV)
pixel size is 90 μm. What magnification is c) Exposure time
required so that diameter of the microcalcifica- d) Focal spot size
tion in the image is twice the pixel size of the
detector?
2.11.4 Case 4: Digital Breast
a) 1
Tomosynthesis: Artifacts due
b) 1.8
c) 2
to High-Contrast Objects
d) 2.3 7. The resolution in digital breast tomosynthesis
images is isotropic, similar to computed
2.11.2 Case 2: Focal Spot Size tomography images.
a) True
Selection in Magnification Views
b) False
3. In magnification imaging, a small focal spot is
used to eliminate which of the following? 8. The origin of the artifacts from high-contrast
a) Scattered radiation objects in digital breast tomosynthesis is most
b) Patient motion similar to:
c) Focal spot blur a) Detector pixel dropout artifact in conven-
d) Geometric distortion tional mammography and radiography
b) Artifacts from metal implants in
4. How wide is the blur from a 0.3-mm focal spot
radiography
when the magnification factor is 2?
c) Streaking artifacts from metal implants in
a) 0.15 mm
computed tomography
b) 0.3 mm
c) 0.6 mm
d) 0.8 mm

40
| 09.11.19 - 01:53

2.11 Review Questions

2.11.5 Case 5: Effect of Image 14. If a small speck-like artifact is observed in the
artifact image during weekly quality control
Post-Processing on the
testing, which actions should be taken?
Appearance of a Mammogram
a) Clean all surfaces
9. The effect of a high-pass filter is to: b) Rotate phantom
a) Increase image resolution c) Gain calibration
b) Emphasize large detail in the image d) Dead pixel mapping
c) Reduce noise e) Replace detector
d) Emphasize small detail in the image

2.11.7 Case 7: Microcalcification-

10. Unsharp masking is a post-processing algo-
rithm that can improve:
like Appearance Caused by a
a) Spatial resolution
Detector Artifact
b) Displayed contrast resolution 15. Which of the following is correct concerning
c) Image noise artifacts from dead pixels in the detector?
d) Geometric distortion a) Dead pixels always appear as black spots.
b) Dead pixels artifacts are always located at
2.11.6 Case 6: Artifact due to identical image coordinates.
Detector Row Dropout c) The detector is free of dead pixels if a
uniform test image does not show any arti-
11. What is a likely cause for a partial detector row
facts.
dropout?
d) An artifact in a uniform test image always
a) Dead pixels are lined up perfectly along
requires the detector to be replaced.
a detector row.
b) The image receptor consists of multiple tiled 16. Which remedial actions should be considered
flat-panel detector elements. when a detector artifact has been identified?
c) The detector electronic readout process is a) Detector replacement
being interrupted. b) Gain calibration
d) The detector panel has been physically c) Dead pixel mapping
damaged. d) All of the above

12. In tomosynthesis, a detector row dropout

artifact looks the same in the synthetic 2D
2.11.8 Case 8: Artifact due to
mammogram as it would in a conventional
Imperfection in Compression
FFDM mammogram. Paddle
a) True 17. What is the most likely cause of a high-contrast
b) False small speck-like image artifact at a fixed loca-
tion, which persists while imaging patients?
13. Prior to replacing the detector, which steps can a) Bad pixels in detector
be taken to resolve a small speck-like image arti- b) Dust during gain calibration
fact? c) Imperfection in the compression paddle
a) None d) Dust on breast support
b) Clean all surfaces
c) Gain calibration 18. Compression force is applied
d) Dead pixel mapping a) To reduce breast thickness
e) All of the above except choice a b) To reduce the potential for patient motion

41
| 09.11.19 - 01:53

Mammography

c) To reduce average glandular dose c) Connection to external battery pack

d) All of the above d) Abandoned cardiac leads

2.11.9 Case 9: Patient Motion Equations

Causing Blurred Parenchymal Object magnification factor M, for an object posi-
Structure in a Mammogram tioned at a distance of dmag from the focal spot, and
19. What is the maximum allowable compression a distance ddet between the focal spot and the
force in mammography? detector:

a) 25 pounds
M ¼ d det =d mag
b) 35 pounds
c) 45 pounds Diameter of the image of an object (xmag), imaged
d) 55 pounds with magnification factor M, when the actual size
of the object is x:
20. Which of the factors below is generally not a
x mag ¼ M  x
cause for patient motion?
a) Breast compression force too low Width of focal spot blur (wblur) for a focal spot of
b) Long exposure time size w:
c) Magnification imaging
d) Use of an X-ray grid w blur ¼ w  ðddet  dmag Þ=dmag
¼ w  M  1Þ

2.11.10 Case 10: EMI Artifact due

to LVAD Device
21. Which of the following statements is correct?
a) EMI artifacts can occur in computed
radiography.
b) EMI artifacts only occur in digital
mammography detectors because of the
high resolution.
c) The appearance of an EMI artifact is similar
to a grid artifact.
d) An EMI artifact is a permanent detector
artifact.
References
[1] Villafana T. Generators, X-ray tubes, and exposure geometry
in mammography. Radiographics. 1990; 10(3):539–554
[2] Pisano ED, Cole EB, Hemminger BM, et al. Image processing
algorithms for digital mammography: a pictorial essay.
Radiographics. 2000; 20(5):1479–1491
[3] Seeram E, Seeram D. Image postprocessing in digital radiolo-
gy—a primer for technologists. J Med Imaging Radiat Sci.
2008; 39(1):23–41
[4] Yaffe MJ, Rowlands JA. X-ray detectors for digital radiography.
Phys Med Biol. 1997; 42(1):1–39
[5] Ayyala RS, Chorlton M, Behrman RH, Kornguth PJ, Slanetz PJ.
Digital mammographic artifacts on full-field systems: what
are they and how do I fix them? Radiographics. 2008; 28(7):
1999–2008

22. EMI artifacts occur due to the

a) Presence of metal in the LVAD
b) Spinning motor in the LVAD

42
| 09.11.19 - 01:53
3 Computed Tomography
Karen L. Brown and Jason R. Gold
Introduction
In computed tomography (CT) imaging, the
X-ray tube and detector array rotate around the
patient in a fixed geometry generating thou-
sands of attenuation measurements through the
patient volume of interest. The attenuation
measurements are reconstructed into axial
images which can then be reformatted into
sagittal and coronal planes. The grayscale value
displayed in each pixel represents the CT number
in Hounsfield units of the corresponding voxel
volume. CT number is a normalized measure-
ment of the linear attenuation coefficient meas-
ured in the voxel relative to that of water,
defined to be zero in value. As such, tissues that
are more attenuating than water will have posi-
tive values in a CT image, and tissues that are less
attenuating than water will have negative values.
Each type of tissue in a CT image has a specific
Hounsfield unit range which may be used quan-
titatively to characterize tissues or pathology.
Image quality in CT is affected by many parame-
ters, some of which are selected by the operator
prior to the acquisition, while others are selected
prior to reconstruction of the data into tomo-
graphic planes. The same acquisition data can,
therefore, be reconstructed in multiple ways to
achieve different image quality end points
dependent upon the clinical goals of the study.
43
| 09.11.19 - 01:53
Computed Tomography
3.1 Case 1: Ring Artifact
3.1.1 Background
● Patient with a history of nephrolithiasis
presented for CT of abdomen and pelvis
examination.
● A helical acquisition was acquired using a
detector configuration of 24 × 1.2 mm channels.
● Axial images of 3.0 mm thickness were
reconstructed using a soft-tissue filter.
3.1.2 Findings
A partial ring artifact is centrally located in all axial
reconstructions.
3.1.3 Discussion
The X-ray tube and detector array rotate around
the CT gantry in a fixed geometry. The detector
array consists of multiple detector channels in
the z-direction, each containing many hundreds
of individual detector elements in the x/y direc-
tion (▶ Fig. 3.1b).1 Each detector within the array
measures the residual X-ray signal through the
patient. If an individual detector within the array
is not properly calibrated with respect to the
Fig. 3.1 (a) Partial ring artifact on helically acquired abdomen–pelvis computed tomography examination.
(b) Multichannel CT detector array. (c) Full ring artifact on axially acquired phantom image.
44
other detectors, or fails completely, artifacts will
appear in the reconstructed images. In helical
acquisitions, the artifact will appear as a partial
ring (▶ Fig. 3.1a) and appear to rotate around
isocenter throughout the imaging volume. Full
ring artifacts will be present in axial scan
acquisitions (▶ Fig. 3.1c). The artifact will be
limited to the images corresponding to the
affected detector channel. For example, if a 16-
channel system has a defective detector in
channel 1, a full ring artifact will appear in the
first reconstructed image and then again in
image 17 but will not appear in images 2 to 16,
assuming the width of the reconstructed slice is
equal to the width of the detector channel.
3.1.4 Resolution
Axial acquisitions of a uniform water phantom,
reconstructed at the thinnest possible slice thick-
ness, should be acquired and evaluated for ring
artifacts by the CT technologist on a daily basis.2
When ring artifacts are identified, some CT
systems will provide the user a method to recali-
brate the detectors (often called an air calibration
scan). If the air calibration scan is not available, or
does not resolve the ring artifact, service
personnel should be contacted and corrective
maintenance be performed.
| 09.11.19 - 01:53

3.2 Case 2: Effect of Patient Size on CT Number Accuracy

3.2 Case 2: Effect of Patient
Size on CT Number Accuracy
3.2.1 Background
● Bariatric patient with a history of renal calculi
presents with abdominal pain.
● A helical abdomen/pelvis CT is performed.
● Incidental finding in the adrenal gland with
elevated CT number measurement.
3.2.2 Findings
Use of CT number for quantitative analysis can be
compromised in bariatric patients due to beam
hardening and truncation artifact.
3.2.3 Discussion
The X-ray tube used in CT produces a polyener-
getic X-ray beam. Filters are placed at the exit
port of the beam to remove low-energy X-rays
increasing the average energy of the X-ray beam.
A process called beam hardening. As the X-ray
beam enters the patient, beam hardening
continues as the lower-energy X-rays in the beam
are attenuated with higher probability. The
displayed CT number is a relative measure of
attenuation as compared to the attenuation of
water. As the beam becomes more energetic
(hardened), less attenuation occurs in a given
tissue which changes the CT number. In a
typically sized patient, this beam hardening effect
is corrected for by the scanner. In bariatric
patients, the beam hardening effect is more
significant and can affect the accuracy of
displayed CT numbers. When the patient’s tissues
fall outside of the scan field of view (FOV), the
system overestimates the attenuation provided
by the tissues within the FOV. This is often
referred to as a truncation artifact and appears as
bright areas in the image (▶ Fig. 3.2a). Truncation
artifact can also affect the accuracy of CT numbers
used for quantitative analysis.3
3.2.4 Resolution
Some systems may offer an extended FOV recon-
struction option as shown in ▶ Fig. 3.2b. In this
case, the visual appearance of the truncation arti-
fact was reduced by reconstructing using an
extended FOV; however, little effect on measured
CT number in the adrenal gland was realized.
Patient positioning can also have a significant
impact on beam hardening and truncation arti-
facts in bariatric patients. The patient in this case,
presented for another scan 2 weeks later
(▶ Fig. 3.2). Note the difference in the patient’s
apparent shape and diameter. This was due to
more effective wrapping of extraneous tissues by
the technologist prior to scanning and is not
related to patient weight loss. No truncation arti-
fact is present and the change in measured Houns-
field units is significant.

Fig. 3.2 (a) Abdomen-pelvis computed tomography showing truncation artifact (arrows) and artificial elevation of
measured CT numbers. Acquisition parameters: 120 kV, 557 mAs. (b) Extended field-of-view reconstruction of image
shown in (a). (c) Follow-up abdomen-pelvis CT of patient. Acquisition parameters: 120 kV, 414 mAs. All other technique
and reconstruction parameters were consistent with image shown in (a).

45
| 09.11.19 - 01:53
Computed Tomography
3.3 Case 3: Effect of kV
Selection on Image Quality and
Dose
3.3.1 Background
● Patient with a history of cecal and appendiceal
adenocarcinoma with multiple prior CT exami-
nations presents for chest/abdomen/pelvis CT
with contrast post resection with peroneal
nodules for evaluation of treatment.
● A helical acquisition is acquired at reduced kV
compared to prior studies.
3.3.2 Findings
Contrast is improved at the lower kV setting at
a dose savings of approximately 17%.
3.3.3 Discussion
Lowering tube voltage (kV) produces an X-ray
beam with lower average and maximum energy.
Lower X-ray energies will be attenuated in a given
tissue more than higher-energy X-rays. This is
primarily governed by differences in attenuation
that occurs due to photoelectric absorption which
is approximately proportional to the atomic
number (Z) cubed of the tissue, and inversely
proportional to the energy (E) cubed of the X-ray
beam. As such, the relative attenuation between
two tissues increases at lower kV settings. On a
CT image, this results in a greater difference in
Hounsfield units between the two tissues, and
therefore, greater contrast. Note the increased
Fig. 3.3 (a) Image acquired at 100 kV and 200 mAs with a displayed volume CT dose index (CTDIVOL) of 8.37 mGy.
(b) Image acquired at 120 kV and 140 mAs with a displayed CTDIVOL of 10.06 mGy.
46
contrast of the seminal vesicle structure shown in
▶ Fig. 3.3a which was acquired at 100 kV as
compared to the same structure in ▶ Fig. 3.3b
acquired at 120 kV.
Tube voltage (kV) also affects the efficiency of
X-ray production in the X-ray tube. At a lower kV
setting, the number of X-rays produced is reduced
by the ratio of the change in kV squared to cubed
(ΔkV2–3). Quantum noise increases when fewer
X-rays are used to make the image, which
will have a negative effect on the visibility of
low-contrast structures. To compensate, a decrease
in kV is often accompanied by an increase in tube
current (mA). As shown in ▶ Fig. 3.3a, b, an appro-
priate increase in mA to achieve an acceptable level
of quantum noise levels can be achieved at a signifi-
cant dose reduction.
3.3.4 Resolution
Tube voltage and current settings should be
adjusted to optimize image quality and dose. The
usefulness of this technique will depend on
patient size and the anatomy/pathology of
interest. A lower kV setting can often be used on
smaller patients. In very large patients, the mA
limitations of the X-ray tube may result in
increased photon starvation artifacts. Beam hard-
ening artifacts will also be enhanced at low
kV settings. Lower-energy beams are also effec-
tive when using iodinated contrast agents as the
average beam energy more closely aligns with
the k-edge absorption peak of iodine. Care should
be taken when using CT numbers for quantitative
analysis as the CT number is affected by the
kV setting.
| 09.11.19 - 01:53
3.4 Case 4: Image Quality Variation with Reconstructed Slice Thickness
3.4 Case 4: Image Quality
Variation with Reconstructed
Slice Thickness
3.4.1 Background
● Patient presents to the emergency department
following a motor vehicle collision.
● CT thorax, CT abdomen, and pelvis examinations
following the administration of contrast were
acquired.
● Axial images of 3.0 and 1.5 mm slice thickness
were reconstructed.
3.4.2 Findings
Small fracture of the clavicle is not visible with
3.0 mm reconstructed slices.
3.4.3 Discussion
Reconstructed slice thickness affects the level of
quantum noise, spatial resolution, and partial
volume averaging present in CT images. Slice
thickness affects the size of the reconstructed
tissue voxel. Each voxel of tissue is displayed as
one shade of gray in the corresponding image
pixel. When the voxel contains more than one type
of tissue or pathology, the CT number displayed
Fig. 3.4 (a) Computed tomography (CT) image reconstructed with 1.5 mm slice thickness. (b) CT image reconstructed
with 3.0 mm slice thickness.
represents the average attenuation measurement
of all the tissues within the voxel. This principle is
called partial volume averaging and is inherent to
the CT image reconstruction process. As the tissue
voxel becomes larger and there is more signal
averaging, there is less spatial differentiation of
the signal in a given direction and the image
appears more blurred, with lower spatial resolu-
tion. ▶ Fig. 3.4a is reconstructed with a 1.5 mm
slice thickness and ▶ Fig. 3.4b is reconstructed
with a 3.0 mm slice thickness. The thinner recon-
structions are sharper (better spatial resolution)
and have less partial volume averaging. Note the
subtle clavicle fracture seen on the 1.5-mm recon-
structions which is not visible on the 3.0-mm
reconstructions. Quantum noise is also affected by
the size of the reconstructed voxel. The amount of
signal (number of X-rays) interacting within a
given voxel is directly proportional to voxel size.
Quantum noise changes inversely with the square
root of the change in signal.
3.4.4 Resolution
Voxel size is a function of the pixel size and the
reconstructed slice thickness selected by the oper-
ator. When reconstructed slice thickness is
increased, more partial volume averaging occurs in
the slice thickness direction. When a thinner
reconstructed slice thickness is selected, there is
better differentiation of tissues along the slice
47
| 09.11.19 - 01:53
Computed Tomography
thickness direction resulting in better spatial reso-
lution and less partial volume averaging. CT proto-
cols often include multiple reconstructions of the
same acquisition at different reconstructed slice
thicknesses to provide the clinician with high
spatial resolution image series as well as low noise
image series for evaluation.
48
The only other parameter that affects the size
of the reconstructed voxel in CT is the FOV
selected by the operator prior to reconstruction.
FOV is often selected to encompass the anatomy
of interest and determines the size of the tissue
voxel and corresponding image pixel in the x and
y dimension.
| 09.11.19 - 01:53
3.5 Case 5: Image Quality Variation with Reconstruction Filter
3.5 Case 5: Image Quality
Variation with Reconstruction
Filter
3.5.1 Background
● Patient with a history of bacterial meningitis and
multiple brain abscesses presents for follow-up
CT to evaluate the response to treatment with
IV antibiotics.
● Helical CT of the head is acquired with contrast
and reconstructed using filtered back projection.
● A reduced mA technique is implemented in an
effort to reduce patient radiation dose.
3.5.2 Findings
The reconstructed images have elevated quantum
noise due to the use of a reduced mA technique.
3.5.3 Discussion
The level of quantum noise in a CT image is related
to 1/√N, where N is equal to the number of X-ray
photons used to generate the image, also called the
signal intensity. The tube current (mA) used during
the acquisition of the exam affects the total number
of X-rays produced. If the mA is reduced by one
half, signal intensity is reduced by one half. They
are directly proportional. The level of quantum
noise in the image will increase as 1/√ (change in
signal) or 1/√1/2 (approximately 40%). In this case,
the images reconstructed using filtered back
projection (▶ Fig. 3.5a) are very noisy. The radiol-
ogist is concerned that important findings may be
missed as increased levels of quantum noise
diminish visibility of low-contrast structures. The
advantage of reduced mA techniques is lower
patient dose. Patient dose is directly proportional to
the mA setting selected by the technologist.
3.5.4 Resolution
There are several acquisition and reconstruction
parameters that affect the level of quantum noise
in an image. One of the reconstruction parameters
selectable by the operator is the reconstruction
filter or kernel. The reconstruction filter will affect
how much smoothing out of the noise occurs in
the image. A soft tissue or standard filter will have
lower image noise compared to a sharp filter, such
as a bone or lung filter, at the expense of lower
Fig. 3.5 (a) Filtered back projection
reconstruction using a smooth filter
with a window width/window level
setting of 100/40. (b) Filtered back
projection reconstruction using
a sharp filter with a window width/
window level setting of 100/40.
(c) Filtered back projection recon-
struction using a sharp filter with a
window width/window level setting of
2500/500. (d) Iterative reconstruction
using a smooth filter with a window
width/window level setting of 100/40.
49
| 09.11.19 - 01:53
Computed Tomography
spatial resolution. The image in ▶ Fig. 3.5a was
reconstructed using a soft-tissue (smooth) recon-
struction filter. At the same window/level setting,
the same data reconstructed with a sharp filter
(▶ Fig. 3.5b) has much higher noise but spatial
resolution has improved. Sharp reconstructions
are often viewed on a different window/level
setting to emphasize the spatial information in the
structures of interest (▶ Fig. 3.5c).
The type of reconstruction algorithm selected by
the operator will also have an effect on image noise.
50
Iterative reconstruction techniques are available on
all modern CT scanners and produce images with
lower noise than filtered back projection techni-
ques.4 ▶ Fig. 3.5d shows the same image recon-
structed using an iterative technique. Note the
significant reduction in quantum noise compared
to ▶ Fig. 3.5a. Spatial resolution in the iterative
reconstruction is also maintained which is another
advantage of this technique. Iterative reconstruc-
tion is often used with reduced mA or kV techni-
ques to overcome increases in quantum noise.
| 09.11.19 - 01:53
3.6 Case 6: Displayed Volume CT Dose Index and Patient Size
3.6 Case 6: Displayed Volume CT
Dose Index and Patient Size
3.6.1 Background
● A 6-year-old patient with medulloblastoma
presents for CT of the thorax with contrast for
evaluation of possible pulmonary embolus.
● Patient has received multiple CT examinations
over the past year.
3.6.2 Findings
Volume CT dose index (CTDIVOL) and dose length
product (DLP) displayed on the dose summary
page underestimate dose to the patient.
3.6.3 Discussion
Patient dose is dependent on the radiation
output of the CT scanner and patient size. The
CTDIVOL (in units of mGy) displayed on the CT
scanner is a measure of the radiation output as
estimated to one of two-sized polymethyl metha-
crylate phantoms, a small 16-cm-diameter
phantom or a large 32-cm-diameter phantom.5
When patient size varies from the size of the
phantom used, the CTDIVOL displayed on the CT
scanner may over- or underestimate patient
dose. In the case presented, the dose summary
report (▶ Fig. 3.6a) estimates 2.08 mGy as the
CTDIVOL to the large 32-cm phantom from the
chest scan. The actual diameter of the patient, as
shown in ▶ Fig. 3.6b is significantly less than
32 cm. In this case, the CTDIVOL underestimates
the dose to the patient which would be higher
than the value displayed on the dose summary
page. When patient size is larger than the
phantom indicated on the dose summary page,
patient dose is overestimated and would be
lower than the displayed CTDIVOL value.
The DLP is calculated by multiplying the esti-
mated CTDIVOL by the scan length and has units of
Fig. 3.6 (a) Computed tomography dose summary page. (b) Axial CT scan used to estimate patient diameter.
mGy-cm. DLP provides an estimate of the total
energy imparted to the scan volume.5 Any error in
estimating patient dose using CTDIVOL is propa-
gated in the calculation of DLP.
The use of dose values displayed on the
CT scanner to estimate patient dose can be particu-
larly problematic for chest, abdomen, and pelvic
examinations of pediatric patients. The displayed
dose for body scans is typically estimated using the
large 32-cm phantom (although some older systems
may use the 16-cm phantom) which may underesti-
mate dose in pediatric patients by a factor of 2 or 3.
For this reason, care must be taken when using
displayed dose metrics for benchmarking and
protocol optimization.
3.6.4 Resolution
When it is necessary to estimate patient dose
from a CT examination, the clinician should be
aware of the limitations of displayed or reported
CT dose metrics on the scanner. The American As-
sociation of Physicists in Medicine has developed
a method to provide a better estimate of patient
dose.6 The size-specific dose estimate (SSDE) is
calculated by multiplying the displayed CTDIVOL
by a correction factor that accounts for the diam-
eter of the patient as compared to the diameter of
the phantom used to estimate CTDIVOL. SSDE is
not currently displayed on CT scanners or on the
dose summary page. Some third-party dose ma-
nagement software programs do provide SSDE
calculations that can be compared to national
benchmarks such as the American College of
Radiology Dose Index Registry. If these tools are
not readily available to the clinician, a medical
physicist should be consulted to provide a patient
dose estimate. As a general rule of thumb, if the
large (32 cm) phantom is used by the scanner to
estimate CTDIVOL, the dose to a 16-cm-diameter
patient will be approximately two times the value
displayed on the scanner. The estimated SSDE for
the patient presented in this case is 3.6 mGy.
51
| 09.11.19 - 01:53
Computed Tomography
3.7 Case 7: Beam Hardening
Artifact
3.7.1 Background
Patient with suspected subdural hematoma
presents for head and neck CTA with and without
contrast.
3.7.2 Findings
Beam hardening artifacts are observed in contrast-
enhanced images.
3.7.3 Discussion
Beam hardening occurs when the polyenergetic
X-ray beam passes through material.7 Higher
density and higher atomic number materials such
as metal, bone, and contrast agents preferentially
attenuate lower-energy X-rays resulting in a
higher average energy X-ray beam exiting the
material. Because, the higher-energy beam has
greater penetrating ability, tissues that lie
between, or are adjacent to these structures
appear to be less attenuating. In the reconstruc-
tion, lower CT numbers than what actually repre-
sent the tissue are calculated and displayed. This
results in shadowing or dark bands as shown in
▶ Fig. 3.7a. The magnitude of beam hardening
that occurs will depend on the path length of the
52
X-ray beam through these dense objects. Note the
presence of beam hardening around the bone
structures in the noncontrast image (▶ Fig. 3.7b).
In the contrast image, beam hardening artifact is
more noticeable when the beam passes through
all three contrast-filled structures but is not
apparent at other angular pathways where the
beam passes through a shorter axis of each indi-
vidual contrast-filled structure.
3.7.4 Resolution
Beam hardening artifacts can be minimized with
the use of higher-energy X-ray beams. Increasing
the tube voltage (kV) produces a higher-energy
X-ray beam but this will also affect image contrast
and patient dose. Dual energy CT acquisitions
acquire data at two separate beam energies. A low
kV setting, such as 80 kV, and a high kV setting, such
as 140 kV, are typical. The attenuation difference
between tissues will be different at the low kV
setting compared to the high kV setting. This infor-
mation can be used to reconstruct virtual monoe-
nergetic images at higher energy, eliminating the
effect of low-energy X-rays on beam hardening.8,9
Angulation of the gantry to avoid passage of the
beam through the long axis of known, high-density
structures will help mitigate the artifact. This
option is not available on all CT scanners. Iterative
reconstruction techniques can also be used to mini-
mize the appearance of beam hardening artifacts.
Fig. 3.7 (a) Contrast scan showing
shading artifact from beam hardening
when the angular orientation of the
beam passes through multiple
contrast-filled structures. (b) Noncon-
trast scan shows less significant beam
hardening artifact when the beam
passes through the long axis of bone
structures.
| 09.11.19 - 01:53
3.8 Case 8: Partial Volume
Artifact
3.8.1 Background
● ICU patient presents with history of right thalamic
hemorrhage and right frontal contusion.
● Portable axial CT scan of the head without
contrast with a reconstructed slice thickness
of 5 mm is performed.
3.8.2 Findings
Hyperdensity in right frontal lobe due to partial
volume averaging is not indicative of contusion
per history.
3.8.3 Discussion
Each pixel in a CT image displays a single shade
of gray representing the calculated CT number
for the associated voxel of tissue. When more
than one type of tissue is contained within a
given voxel, the average attenuation value of the
tissues within the voxel is used to calculate the
CT number and corresponding displayed gray-
scale value. Head CT scans are commonly recon-
structed with a 5 mm slice thickness, resulting in
relatively large voxels in the z-direction (through
the scanner bore). ▶ Fig. 3.8a shows a hyperden-
sity in the right frontal lobe region indicated by
Fig. 3.8 (a) Axial computed tomography of the brain reconstructed with 5 mm slices shows hyperdensity in the right
frontal lobe. (b) Scout image showing location of adjacent axial slices (white lines) and voxel encompassing both skull
and brain matter. Note: for illustrative purposes only, voxel size and position are not accurately depicted. (c) Axial follow-
up CT shows no hyperdensity in the right frontal lobe indicating the effect on prior image was likely due to partial
volume averaging.
3.8 Case 8: Partial Volume Artifact
the arrow. The increased signal is due to the
averaging of skull and brain matter attenuation
within the voxel (▶ Fig. 3.8b) resulting in
elevated CT number and brighter shade of gray
presentation in the image. Follow-up CT
(▶ Fig. 3.8c) shows no hyperdensity due to
contusion and that the elevated CT numbers are
likely due to partial volume averaging.
3.8.4 Resolution
The magnitude of partial volume artifact is
primarily affected by reconstructed slice thickness.
Slice thickness controls the z-dimension of a tissue
voxel. It is common for most CT examinations to
be reconstructed using thick and thin reconstruc-
tions. Partial volume averaging will be diminished
in thinner slice images compared to thick image
reconstructions. The x–y dimension of the voxel is
equal to the size of the image pixel which is deter-
mined by the FOV divided by the size of the image
matrix. For most modern CT scanners, the matrix
size is fixed at 512 × 512 pixels. Adjusting the FOV
is, therefore, the single parameter adjustment that
can affect pixel size. The size of each pixel in CT is
typically much smaller than the slice thickness, so
changing FOV has less effect on partial volume
averaging than slice thickness. Reconstructing and
viewing images in different reconstruction planes
may also affect partial volume averaging as the
tissue types within a given voxel may vary along
the direction of reconstruction.
53
| 09.11.19 - 01:53

Computed Tomography

3.9 Case 9: Metal Artifact 3.9.4 Resolution

3.9.1 Background
● Patient with possible leg cellulitis and fluid collec-
tion in the anterolateral leg is referred for CT.
● Helical CT scan of the lower leg with 1.0- and
3.0-mm reconstructions is acquired.
3.9.2 Findings
Metal artifact precludes visualization of possible
soft tissue fluid collection.
3.9.3 Discussion
The presence of metal within the CT scan FOV
causes severe streaking artifact. There are
several phenomenon contributing to the appear-
ance of streaking including beam hardening and
photon starvation effects. In addition, the meas-
ured linear attenuation values of metal struc-
tures may be beyond the dynamic range of the
CT system. Visualization of anatomy and path-
ology in the vicinity of metal material may be
significantly compromised as a result. In this
case, the radiologist was unable to make a defin-
itive diagnosis due to the severe streaking arti-
fact (▶ Fig. 3.9a, b) in proximity to the structures
of interest.
There are several techniques to minimize the
appearance of metal artifacts.11 Thin acquisition
sections are recommended to reduce partial
volume averaging; however, larger reconstructed
slice thickness averages the noise over larger
voxels and mitigates the appearance of streaking.
▶ Fig. 3.9a was reconstructed with 3.0 mm slice
thickness and shows mild reduction in streak
artifact compared to ▶ Fig. 3.9b which was
reconstructed with 1.0 mm slice thickness.
Higher tube voltage (kV) increases the pene-
trating ability of the X-ray beam through metal
structures and also increases the number of
X-ray photons in the beam helping to overcome
both beam hardening and photon starvation
effects. Note significant reduction in metal
streak artifact seen in ▶ Fig. 3.9c acquired at
140 kV. Dual energy CT can be used to generate
virtual monoenergetic images.89 Using this tech-
nique, low-energy X-rays in the beam that
contribute significantly to beam hardening are
removed, minimizing their contribution to
streak artifacts. Metal artifact reduction soft-
ware is available from equipment and third party
vendors. The use of iterative reconstruction
techniques reduces noise, minimizing artifact
intensity. These techniques can be used sepa-
rately or in combination to reduce the appear-
ance of metal streak artifacts.

Fig. 3.9 (a) Metal artifact reconstructed with 3.0 mm slice thickness showed mild improvement of streak artifact
compared to 1.0 mm reconstruction as shown in (b). (b) Image reconstructed with 1.0 mm slice thickness shows slightly
enhanced streaking from metal artifact compared to 3.0 mm reconstruction as shown in (a). (c) Image acquired at
140 kV shows significant reduction in metal streak artifact.

54
| 09.11.19 - 01:53
3.10 Case 10: Motion Artifact
3.10.1 Background
● Patient presents with left-sided weakness.
● Noncontrast helical CT is performed with sagittal
and coronal reformats.
3.10.2 Findings
Evaluation of the study was compromised due
to motion artifact, requiring the technologist to
repeat the scan.
3.10.3 Discussion
The appearance of motion artifacts can be quite
varied depending on the source of motion (patient,
respiratory, cardiac, etc.) and severity. Respiratory
motion may appear as a generalized loss of resolu-
tion in the anterior chest cavity with little effect in
other areas of the image. As shown in ▶ Fig. 3.10,
significant patient motion can cause streaking,
shading, ghosting, and incongruence of anatomical
features. In some cases, the degradation in image
quality caused by motion requires the scan to be
repeated resulting in increased radiation dose to
the patient.
3.10.4 Resolution
Patient motion can be mitigated with adequate
exam preparations to include careful patient posi-
tioning and appropriate use of immobilization
devices, explanation of the examination process to
the patient prior to the scan, and provision of clear
instructions throughout the scanning process. Opti-
mization of scan parameters is also essential, when
imaging anatomical features with higher probability
of involuntary motion, gantry rotation time and
helical pitch are adjusted accordingly to optimize
scan acquisition time. Shorter gantry rotation times
and higher helical pitch result in shorter scan
acquisition times. For example, head scans typically
3.10 Case 10: Motion Artifact
Fig. 3.10 Image presentation of significant patient
motion during computed tomography acquisition.
incorporate longer gantry rotation times (0.8–2.0
seconds) as compared to abdominal scans (0.25–0.5
seconds) during which respiratory motion is more
likely to occur. For patients with a high probability
of motion, either due to their condition or age, seda-
tion may be appropriate. Systems with larger beam
widths and dual source technology also provide
opportunities for faster scanning, minimizing the
potential for motion artifacts.
Cardiac motion creates a significant challenge
when imaging heart structures and vasculature.
Artifacts related to cardiac motion can manifest in
a variety of forms including blurring, ghosting, and
misregistration.12 Prospective or retrospective
cardiac gating techniques are often employed to
essentially “freeze” heart motion during the
selected portion of the cardiac cycle. The effective-
ness of these techniques is dependent on patient
heart rate and stability. Beta blockers may be
administered to the patient to decrease patient
heart rate and extend the period of diastole when
using cardiac gating techniques.
55
| 09.11.19 - 01:53

Computed Tomography

3.11 Review Questions b) Bone

c) Fat
3.11.1 Case 1: Ring Artifact d) Iodine
1. What reconstructed slice thickness will provide
the greatest visibility of detector malfunctions 6. What CT acquisition parameter has the largest
in a CT scanner? effect on quantum noise in the image?
a) 0.6 mm a) mA
b) 1.2 mm b) kV
c) 3.0 mm c) Rotation time
d) 5.0 mm d) Collimation

2. What is the purpose of the daily air calibration 3.11.4 Case 4: Image Quality
scan recommended by some CT scanner manu- Variation with Reconstructed Slice
facturers? Thickness
a) Assess the CT number accuracy of water and
7. By what factor does quantum noise in the
air
image change when slice thickness is decreased
b) Adjust the gain settings of individual
from 3.0 to 1.5 mm?
detector elements
a) 0.75
c) Measure noise standard deviation of the
b) 1.0
system
c) 1.4
d) Disable malfunctioning detector channels
d) 2.0

3.11.2 Case 2: Effect of Patient

8. What is the primary advantage of thicker
Size on CT Number Accuracy reconstructed slice thickness in CT?
3. What factor affects the measured CT number a) Improved spatial resolution
for a given tissue? b) Faster acquisition times
a) Beam energy c) Lower patient dose
b) Tube current modulation d) Increased low-contrast visibility
c) Gantry rotation speed
d) Window level/width 3.11.5 Case 5: Image Quality
Variation with Reconstruction
4. What affect does increasing the FOV have on
Filter
image spatial resolution?
a) Decreases 9. An increase in what parameter will reduce
b) Increases quantum noise in a CT image with no direct
c) Stays the same effect on patient dose?
a) Pitch
b) Slice thickness
3.11.3 Case 3: Effect of kV
c) Rotation time
Selection on Image Quality
d) kV
and Dose
5. For a given X-ray beam energy, what material 10. What is the primary advantage of iterative
has the highest probability of attenuating an reconstruction techniques as compared to
X-ray photon? filtered back projection?
a) Air a) Lower noise images

56
| 09.11.19 - 01:53

3.11 Review Questions

b) Faster reconstruction time 3.11.8 Case 8: Partial Volume

c) Reduced image storage requirements
Artifact
d) Minimal partial volume averaging
15. What effect will decreasing FOV have on
quantum noise when using fixed acquisition
3.11.6 Case 6: Displayed Volume
parameters?
CT Dose Index and Patient Size a) Increase
11. How is the displayed CTDIVOL related to the tube b) Decrease
current setting during a CT acquisition? c) No effect
a) Directly
b) Inversely 16. What effect will decreasing reconstructed slice
c) Exponentially thickness have on spatial resolution?
d) One over the square root a) Increase
b) Decrease
12. What is a typical CTDIVOL reported for a routine c) No effect
adult abdomen/pelvis CT examination of
average-sized patient?
3.11.9 Case 9: Metal Artifact
a) 1–5 mGy
b) 10–15 mGy 17. What is the dominant interaction of 80 kV
c) 25–30 mGy X-rays in metal?
d) 45–50 mGy a) Coherent scattering
b) Compton scattering
c) Pair production
3.11.7 Case 7: Beam Hardening
d) Photoelectric absorption
Artifact
13. What component of CT imaging system 18. What quantity is used to calculate CT number
corrects for variations in X-ray beam intensity (Hounsfield unit)?
due to different X-ray path lengths through a) Linear attenuation coefficient
the patient? b) Mass attenuation coefficient
a) Anode c) Tissue atomic number
b) Bow-tie filter d) Tissue density
c) Collimator
d) Grid
3.11.10 Case 10: Motion Artifact
14. An increase in what parameter will reduce the 19. What is the effect of using a shorter gantry
difference in grayscale values between adjacent rotation time, assuming all other protocol
tissues in a CT image. parameters remain unchanged?
a) mA a) Decreased spatial resolution
b) kV b) Longer scan acquisition time
c) Rotation time c) Increased quantum noise
d) Pitch d) Higher patient dose

57
| 09.11.19 - 01:53

Computed Tomography

20. What is the effect of increasing helical pitch [3] Fursevich DM, LiMarzi GM, O’Dell MC, Hernandez MA, Sensa-
kovic WF. Bariatric CT imaging: challenges and solutions.
assuming all other protocol parameters remain Radiographics. 2016; 36(4):1076–1086
unchanged? [4] Padole A, Ali Khawaja RD, Kalra MK, Singh S. CT radiation
dose and iterative reconstruction techniques. AJR Am J
a) Improved visibility of low-contrast struc-
Roentgenol. 2015; 204(4):W384–92
tures [5] Report No AAPM. 96, The Measurement, Reporting, and
Management of CT Dose, The American Association of Physi-
b) Lower patient dose
cists in Medicine, 2008
c) Higher spatial resolution [6] Report No AAPM. 204, Size Specific Dose Estimates (SSDE) in
d) Decreased quantum noise Pediatric and Adult Body CT Examinations, The American
Association of Physicists in Medicine, 2011
[7] Barrett JF, Keat N. Artifacts in CT: recognition and avoidance.

Equations Radiographics. 2004; 24(6):1679–1691

[8] McCollough CH, Leng S, Yu L, Fletcher JG. Dual- and multi-en-
ergy CT: principles, technical approaches, and clinical appli-
FOV
Pixel size ¼ cations. Radiology. 2015; 276(3):637–653
Matrix Size [9] Grajo JR, Patino M, Prochowski A, Sahani D. Dual energy
CT in practice: basic principles and applications. Appl Radiol.
Voxel size ¼ pixel size  slice thickness 2016; 45(7):6–12
[10] Katsura M, Sato J, Akahane M, Kunimatsu A, Abe O. Current
 t  w and novel techniques for metal artifact reduction at CT:
CT # ðHUÞ ¼  1000
w practical guide for radiologists. Radiographics. 2018; 38(2):
450–461
μt is the average linear attenuation coefficient of [11] Lee MJ, Kim S, Lee SA, et al. Overcoming artifacts from
metallic orthopedic implants at high-field-strength MR
the tissues within a voxel. imaging and multi-detector CT. Radiographics. 2007; 27(3):
μw is the linear attenuation coefficient of water 791–803
[12] Kalisz K, Buethe J, Saboo SS, Abbara S, Halliburton S, Rajiah P.
Artifacts at cardiac CT: physics and solutions. Radiographics.
2016; 36(7):2064–2083
References
[1] Bushberg JT, Seibert JA, Leidholdt E, Boone J. The Essential
Physics of Medical Imaging. 3rd ed. Philadelphia: Wolters
Kluwer Health/Lippincott Williams & Wilkins; 2012
[2] The American College of Radiology. Computed Tomography:
Quality Control Manual, 2017

58
| 09.11.19 - 01:53
4 Magnetic Resonance Imaging
Puneet Sharma
Introduction
From its inception in the 1970s, and through
humble beginnings in the early 1980s, magnetic
resonance imaging (MRI) has experienced over-
whelming growth in utility and innovation across
virtually all diagnostic applications. The power of
MRI lies in image contrast, exploiting the inherent
magnetic properties of protons predominantly
found in tissue water, but also in other biochemical
species. The essential tool for an MRI experiment
is a large main magnetic field (> 0.5 Tesla), which
creates an observable net magnetization in biolog-
ical samples that can be subsequently manipulated
for image formation. A carefully designed collec-
tion of radiofrequencies (RF) and time-varying
gradient pulses work in conjunction with the main
magnetic field to produce image contrast “maps”
from received signals that are “weighted” toward
specific tissue properties, such as T1, T2, T2*,
proton density (PD), and, if desired, diffusion and
flow. This image contrast versatility is akin to
histopathological tissue staining, wherein specific
tissue substructures are highlighted by altering the
chemical fixation. Anatomical visualization in MRI
can be prescribed in any orientation using multi-
planar 2D or 3D slices; however, gating and repeti-
tion can be employed to manifest a temporal
dimension, allowing cine or 4D visualization.
Since MRI acquisitions depend on tissue relaxa-
tion properties (i.e., T1 and T2), adequate time is
needed to impart appropriate weighting to the
encoded MR signal. In its basic sense, the received
signal represents information pertaining to tissue
properties, pulse timing, and signal location,
which is subsequently stored. A multitude of addi-
tional signals are generated, encoded, and stored
that are eventually combined and decoded
through a process called Fourier analysis to
produce a final image. Though some MR acquisi-
tions are on the order of 500 miliseconds for one
slice, most scan times are on the order of minutes
due to the repeated process of excitation,
encoding, timing, and signal reception. Many
significant innovations in MRI over the past two
decades have focused on reducing acquisition
times, mostly by limiting the required data needed
for image reconstruction.
In actuality, only a tiny percentage of protons
make up the observable net magnetization for
MRI. This places significant importance on coil
reception sensitivity to optimize signal-to-noise
ratio (SNR). Moreover, these SNR constraints
may place practical limits on image resolution,
especially in comparison to computed tomog-
raphy (CT). Although SNR can be improved
through signal averaging, this prolongs scan
times significantly.
Common Image Quality
Problems
● Motion (bulk tissue movement and flow)
● Susceptibility
● Aliasing
● Truncation
● Field inhomogeneity
● Coil use and placement
59
| 09.11.19 - 01:53

Magnetic Resonance Imaging

4.1 Case 1: Appearance of 4.1.3 Discussion

Discrete Image Ghosts on
Abdominal Imaging
4.1.1 Background
● Subject underwent routine abdominal MRI,
using multislice gradient echo imaging.
● This type of gradient echo data acquisition inter-
leaves collection of phase-encoded data signals
from all slices over the course of one repetition
time (TR) period (~ 170 miliseconds).
Subsequent phase-encode steps are repeated
every TR period, until k-space is filled.
● Most abdominal MRI acquisitions, such as this,
require the subject to suspend breathing for the
duration of the scan. Typically, the breath-hold
duration is < 20 seconds.
The essence of MR signal acquisition and recon-
struction relies on data consistency in k-space over
the complete duration of the scan (seconds to
minutes). A positional change of an object during
acquisition will induce an amplitude or phase mo-
dulation of the expected k-space encoding step,
especially in relation to previous (and future)
encoding steps. This modulation in k-space will
manifest as a replication of the object in the resul-
tant image due to properties of the discrete Fourier
transform. The nature of the replication depends on
the nature of the object motion, and how it

4.1.2 Findings
● Subject was not able to suspend breathing
during the acquisition.
● Upon reconstruction, distinct image “copies”
(ghosts) are seen propagating in the anterior-
posterior direction (▶ Fig. 4.1, ▶ Fig. 4.2, and
▶ Fig. 4.3).
Fig. 4.1 Axial T1 gradient echo of the abdomen
showing gross appearance of image ghosts,
propagating in the anterior-posterior direction.

Fig. 4.2 (a–c) Three abdominal cases, showing acquisition during (a) heavy breathing (period motion); (b) incomplete or
irregular breath holding, and (c) perfect breath holding. Variation of phase encode data collection accompanies each case.

60
| 09.11.19 - 01:53
4.1 Case 1: Appearance of Discrete Image Ghosts on Abdominal Imaging
Fig. 4.3 Discrete ghosts from pulsatile flow in the aorta
propagate in the anterior-posterior direction.
synchronizes with data acquisition: periodic
motion, such as respiration, will usually result in
discrete ghosts, while random motion, such as eye
movement and swallowing, will result in faint,
unstructured ghosts.
Since the effective sampling rate is slower in the
phase encode direction (1/TR, for single-echo
imaging; 1/echo-spacing, for echo train imaging)
compared to the frequency encode direction
(1/Δt = bandwidth (BW), Δt = sampling interval), the
sensitivity of k-space modulation and inconsistency
is commonly seen along the phase encode direc-
tion. This is depicted in ▶ Fig. 4.2 which plots the
degree of (normalized) k-space modulation as a
function of three types of motional behaviors: deep
periodic breathing, shallow breathing, and perfect
breath holding. As seen, the change in amplitude
among phase-encode steps are much greater than
individual frequency-encode steps.
Flow due to moving spins also may induce
image ghosts. Like respiratory motion, flow pulsa-
tility coupled with signal amplitude changes from
unsaturated spins moving through the imaging
plane and can cause modulation of k-space across
phase-encode steps (▶ Fig. 4.3).
4.1.4 Resolution
The simplest tactic to remedy the clinical signifi-
cance of image ghosts is to exchange phase and
frequency directions in order to redirect ghosts
into a perpendicular direction, and better reveal
tissue under examination. However, any strategy
to reduce or entirely eliminate image ghosts
involves ensuring k-space consistency over the
duration of the data acquisition. As seen in
▶ Fig. 4.2, eliminating ghosts ultimately requires
synchronizing phase-encode data collection with
known object motion. This synchronization, as
employed with navigator or respiratory gating,
collects data only from particular motional states
(e.g., expiration) and disregards others. Depending
on the complexity of motion, data synchronization
may result in long scan times or even insignificant
motion compensation. Alternatively, if one is able
to reduce TR such that the phase encode sampling
rate is high relative to the object motional rate,
ghosts may become indistinguishable from edge
blurring. Reducing TR and echo spacing may also
allow breath holding. This latter strategy is the
most robust compensation method and is
becoming increasingly more applicable due to
advancements in parallel imaging and compressed
sensing.
61
| 09.11.19 - 01:53

Magnetic Resonance Imaging

4.2 Case 2: A Well-Defined Area
of Signal Hyperintensity Appears
Bilaterally at the Level of the
Internal Auditory Canal on
Diffusion-Weighted MRI,
Affecting Visualization of
Surrounding Structures
4.2.1 Background
● Routine brain MRI exam without specific
pathologic indication. Diffusion-weighted imaging
(DWI) is a standard-of-care acquisition for
describing cellular integrity of various tissue types.
● DWI is an echo-planar imaging (EPI) technique,
utilizing large directionally sensitive gradients to
encode movements related to diffusion.
● EPI is a rapid gradient echo method that collects
all k-space data following one RF excitation
(“single-shot”). This typically involves fast
gradient reversals, interleaved with incremental
phase-encoding steps.
common outcomes: (1) image distortion (or
warping), and (2) susceptibility related signal loss.
When adjacent tissues have very different magnetic
susceptibility, image distortion is likely since the
local magnetic field is altered from its expected
value. This inhomogeneous environment may
disrupt the applied spatial encoding gradients
during an MRI acquisition, which are assumed to be
linear. In the frequency-encoding direction, a
nonlinear gradient will change the overall fre-
quency distribution of encoded spins, where some
spatial locations may now be “mapped” with more
than one spatial frequency (▶ Fig. 4.5), resulting in
image distortion or signal “pile-up.” The more
shallow the applied gradient (low encoding BW),
the more significant the spatial mismapping and
distortion. Even though this effect occurs predomi-
nantly in the frequency-encoding direction, single-
shot EPI shows susceptibility-related distortion in
the phase-encode direction (▶ Fig. 4.4) due to the
relatively low sampling rate in this direction. Accel-
erated signal loss is another result of high magnetic
susceptibility environments. The large local field

4.2.2 Findings
● The high signal intensity is an artifact evolving
from constructive signal “pile-up” related to
susceptibility (arrows, ▶ Fig. 4.4).
● The internal auditory canal (IAC) region is an
air-filled region with changing geometry that
represents a sharp transition of magnetic suscept-
ibility compared to the rest of the brain structure.

4.2.3 Discussion
Tissues and other substances in the body will alter
the applied magnetic field based on their tissue
properties and chemical composition. “Magnetic
susceptibility” is a term that describes the degree to
which a substance is able to disrupt the local
magnetic field in terms of its strengthening (para-
magnetic) or weakening (diamagnetic). Some
substances, such as iron, are ferromagnetic which Fig. 4.4 Diffusion EPI brain acquisition showing hyper-
cause significant disruption of the local magnetic intense signal at the level of the internal auditory canal
caused by susceptibility.
field. Significant magnetic susceptibility has two

62
| 09.11.19 - 01:53
Fig. 4.6 Comparison of signal loss and distortion due to the presence of metal in three types of imaging techniques:
(a) turbo spin echo, (b) gradient echo, and (c) echo-planar imaging.
alterations induce additional spin dephasing in the
transverse plane causing signal loss. Even though
the additional field inhomogeneities are mitigated
by RF rephasing in spin echo techniques, accelerated
T2 decay will still occur. This is not the case for
gradient echo techniques, where T2*-related signal
loss is the predominant result in high magnetic
susceptibility environments (▶ Fig. 4.6). It is impor-
tant to note that exploiting susceptibility effects to
enhance tissue characterization has recently
become an active area of research. Susceptibility-
weight imaging (SWI) uses the effect of suscepti-
4.2 Case 2
Fig. 4.5 (a,b) Frequency encoding
gradient showing gradient warping
due to susceptibility.
bility to characterize tissue properties and is espe-
cially useful for looking at iron content and hemor-
rhages in the brain.
4.2.4 Resolution
Complete elimination of image distortion and
signal loss due to large magnetic susceptibility
may not be attainable, especially for metal
implants or other ferromagnetic substances.
However, there are several tactics to reduce the
impact on surrounding anatomy. Susceptibility
63
| 09.11.19 - 01:53
Magnetic Resonance Imaging
artifacts are most prevalent on gradient echo
sequences and high field strengths (3T), so
switching (if possible) to spin echo alternatives
and lower field strengths (1.5T) are the primary
options. Also, one should address the key parame-
ters that affect signal loss and image distortion.
The former can best be impacted by using lower
times to echo (TEs), while still maintaining the
desired image contrast. Distortion can be
64
addressed by increasing the imaging BW, which
allows for more encoding frequencies and lessens
the degree of spatial misregistration. This also
allows for shorter TEs. While other strategies
exist (such as reducing voxel size), new MR inno-
vations have been developed to incorporate
specific susceptibility-reducing attributes, espe-
cially for imaging in the presence of metal
implants for orthopedic applications.
| 09.11.19 - 01:53
4.3 Case 3: Appearance of Extra Field-of-view Anatomy
4.3 Case 3: Appearance of
Extra Field-of-view Anatomy
on the Inferior Portion of
Sagittal 3D T2-Weighted
Acquisition of the Spine
4.3.1 Background
● 3D MR acquisitions are subject to phase
encoding in two directions.
● Isotropic resolution allows high-resolution
reconstructions in secondary directions.
● Additional image quality issues exist for 3D
acquisitions.
4.3.2 Findings
● Large superior-inferior FOV requires activation
of both multichannel head coil and cervical/
thoracic spine coil elements. Lumbar spine coil
elements are deactivated.
● Ghost-like artifact resembling the head
appears superimposed on the lower part of
the image and on the resultant axial
reconstruction.
● No other abnormal artifacts appear on the upper
half of the sagittal image, nor in the left-right
direction of the axial series.
● 3D T2 TSE sequence was applied using volume
selective RF excitation, but with minimal over-
sampling of data in the in-plane phase-encoding
direction (superior-inferior direction) to save
imaging time.
4.3.3 Discussion
The example shows clear evidence of anatomical
wrap-around artifact, also known as “aliasing,” in
the superior-inferior direction with a section of the
head superimposed onto the lower thoracic spine.
This is also apparent on an axial reformatted slice
(▶ Fig. 4.7a). Aliasing occurs in the phase-encoding
direction for both 2D and 3D acquisitions when
employing Cartesian k-space trajectories. The possi-
bility of aliasing in any imaging scenario depends on
three main situations: (1) whether tissue is within
the sensitivity region of an activated RF receiver coil;
(2) whether the prescribed FOV is smaller than the
anatomical extent in the (2D or 3D) phase-encoding
direction; and (3) the excitation region of the RF
transmission pulse. In ▶ Fig. 4.7, the top of the head
is within a region that has both been excited by an
RF pulse and observable by the activated head coil.
Since the in-plane phase-encode direction is
superior-inferior and the FOV is smaller than this
“activated” anatomy, the portion of the head outside
the FOV is superimposed, or “wrapped,” onto the
other side. This is another form of spatial mismap-
ping due to the MR signal encoding process. To
understand the phenomena, note that the tissue
outside of the prescribed 2D FOV has still been
excited by an RF pulse and thus still subject to
applied field encoding gradients. The phase
encoding process imparts phase shifts between −180
and + 180 degrees within the FOV. However, tissues
outside of the FOV still experience phase encoding,
but with shifts outside this range (> + 180, or < −180).
For example, due to the cyclical nature (i.e., sine
wave) of MR signals, a phase shift of + 185 degrees is
equivalent to −175 degrees and is “wrapped” onto
other encoded steps of −175 degrees as illustrated
in ▶ Fig. 4.8. Aliasing can also occur in the slice
Fig. 4.7 (a) 3D sagittal T2, and
(b) a corresponding axial reformatted
slice of the spine showing faint signs
of head anatomy, indicative of signal
aliasing.
65
| 09.11.19 - 01:53
Magnetic Resonance Imaging
direction in 3D imaging. However, the case in
▶ Fig. 4.7 employed slab-selective RF, so that little to
no tissue outside the intended volume (FOV in the
slice direction) is excited. Even with slab-selective
excitation, some residual aliasing may occur, as
shown in ▶ Fig. 4.9, since no selective RF pulses have
a perfect slice profile.
4.3.4 Resolution
The straightforward remedy for aliasing is to be
conscious of the three criteria mentioned above.
One must first observe what tissue resides outside
the imaging FOV, but within the RF excitation and
activated coil sensitivity region. If possible, specific
coil elements should be deactivated to avoid
encoding this residual signal. However, no receive
coils have sharp sensitivity cut-offs. Moreover, the
66
Fig. 4.8 Graphical representation of
phase-encode wrap around. Excited
anatomy outside the field of view
“acquires” the same phase-encoding
step as corresponding tissue within the
FOV.
Fig. 4.9 (a, b) Slice encode aliasing
shown on axial 3D T2 imaging. Spins
excited outside the prescribed imaging
volume are encoded similarly as those
within the imaging volume. In some
cases, 3D aliasing may be mistaken for
pathology.
entire coil-sensitivity region may be needed for
particular 3D acquisitions. Mostly, users are
resigned to “oversample” the tissue outside the
FOV, especially if one does not want to increase
FOV or coverage for resolution or data limit
purposes, respectively. While this strategy is liber-
ally applied in the 2D phase-encode direction, a
lesser amount is generally applied in the slice-
encoding direction, particularly if slab-selective
excitation is used. Phase encode oversampling
costs time, but adds SNR, and therefore, should be
balanced against other imaging criteria for specific
applications. It should be noted, finally, that some
aliasing is tolerable: if aliased anatomy does not
impinge on the diagnostic region of interest,
significant time-savings or optimized resolution
can be achieved. Two applications that exploit this
are cardiac and phase-contrast imaging.
| 09.11.19 - 01:53
4.4 Case 4: Precontrast, Axial
3D T1-Weighted Gradient Echo
with Fat Suppression Shows
Adequate Anatomical Detail,
but Minor Edge Ripple and Blur
that is Presumed to be Motion
4.4.1 Background
● 3D gradient echo imaging of the abdomen is
a breath-hold technique that will be subject
to motion artifacts if the patient cannot comply.
● Imaging resolution is 1.4 × 1.7 mm in-plane,
with 3 mm (interpolated) slice thickness.
● The subject appeared cooperative with breath
holding instructions, judging from other
acquisitions (not shown).
4.4.2 Findings
● There are several areas of edge ripple and blur in
the image, including the liver capsule and portal
vein.
● There is also some signal fluctuation in the
visceral fat in the retroperitoneum.
● While most ripples emanate in the anterior-
posterior direction, some extend in lateral
directions as well.
● Incomplete breath holding is a source of minor
blur in the phase-encode direction; however,
the external abdominal wall appears fairly sharp,
without indication of signal propagation
(▶ Fig. 4.10).
4.4.3 Discussion
Given the requirements of breath-hold imaging in
the abdomen, 3D acquisitions are significantly
4.4 Case 4
undersampled in both the phase- and slice-
encoding direction to achieve appropriate scan
durations. This amounts to a reduced phase reso-
lution and increased slice thickness, respectively.
When resolution is too low in a certain direction,
truncation (or Gibbs) artifact may occur. Trunca-
tion artifact is identified by periodic low- and
high-signal intensity ripples emanating from
high-contrast edges. While this appearance may
closely mimic motion artifact, truncation effects
are exclusive to sharp, high-contrast edges, and
fade thereafter; there is no replicating of anatomy.
This phenomenon is due to the inability of the
acquisition to accurately define sharp edges, using
the available frequency encoding range, particu-
larly high-frequency data. The consequence is an
overestimation and underestimation of the high-
contrast structure, which decays with distance
(▶ Fig. 4.11). The propagation distance of ripples is
longer if more high-frequency data points are
absent from the acquisition.
In-plane truncation artifacts affect both 2D
and 3D acquisitions. However, truncation in the
slice direction is unique to 3D imaging.
▶ Fig. 4.12 shows a coronal reconstruction of the
original data set, given in ▶ Fig. 4.12. The recon-
structed data makes the truncation artifact more
Fig. 4.10 Breath-hold 3D T1 spoiled-gradient echo with
fat suppression, showing adequate breath holding, but
subtle ringing and edge enhancements (arrows).
Fig. 4.11 Gibbs ripple resulting from
a limited frequency-encoding range.
67
| 09.11.19 - 01:53
Magnetic Resonance Imaging
conspicuous, especially realizing that the axial
3D volume was acquired with 6 mm (3 mm inter-
polated) slice thickness. The ripples, which are
best seen in the spleen and against the fat-
suppressed retroperitoneum, are large due to the
low resolution in the slice direction. The signifi-
cance of this artifact is not appreciated when
viewed in the standard axial orientation.
Fig. 4.12 Coronal reconstruction of 3D data set from
▶ Fig. 4.1, showing more prominent truncation artifact
due to large slice thickness.
68
4.4.4 Resolution
The 3D truncation artifact seen in ▶ Fig. 4.10 can
be improved by reducing the slice thickness, or
rather, improving the slice resolution. Similarly, a
fully sampled in-plane data set, such as
288 × 288, will also reduce edge ripple in 2D.
However, it is important to note that a complete
elimination of edge ringing is not possible. This is
due to the discrete nature of digitized data
sampling in MRI and subsequent Fourier recon-
struction. In practice, an infinite frequency range
would be necessary to describe a sharp step
change in signal intensity between two objects,
which is not possible in MRI experiments.
However, when high enough resolution imaging
is employed, a greater frequency range is avail-
able to approximate edge information, with high-
contrast features being more well defined
(▶ Fig. 4.13). Even though ringing may still be
present, both intensity and ripple distance from
the edge location will be progressively reduced,
which makes them less conspicuous. If increasing
the resolution is not practical due to SNR and
time constraints, applying a filter to smooth the
image helps to reduce much of the ringing, at the
expense of some edge blurring.
Fig. 4.13 Comparison of low- and
high-resolution imaging of a grid
phantom. The degree of truncation
ripple is less in the high-resolution
scan.
| 09.11.19 - 01:53
4.5 Case 5: Dark Etching Appears at the Boundary of Fat and Soft-Tissue Layer
4.5 Case 5: Dark Etching
Appears at the Boundary of Fat
and Soft-Tissue Layers
4.5.1 Background
● Routine Ax T1 TSE for neck soft tissues with
phase encoding left-to-right.
● Subject is cooperative, and slight blur around the
tongue is considered normal.
4.5.2 Findings
● While some dark signal is attributable to signal
loss, specific etching in the locations indicated
seems abnormal (▶ Fig. 4.14).
● Parameters are typical for T1 imaging
(TR ~ 500 milliseconds; TE ~ 11 milliseconds),
with a matrix of 256 and bandwidth less than
200 Hz/pixel.
4.5.3 Discussion
The appearance of dark etching, also known as
“India ink,” between soft tissues and fat is indica-
tive of type 1 chemical shift artifact. Type 1 arti-
fact, which is considered a misregistration arti-
fact between fat and water spins, affects all
conventional imaging techniques such as spin
echo, TSE, and gradient echo. Chemical shift is
somewhat similar to susceptibility-induced
mismapping: when two neighboring tissue
proton species have different resonant frequen-
cies, they may be spatially misinterpreted during
frequency encoding, and therefore spatially offset
from one another (or superimpose) during image
formation. This is the case for fat and water
protons, where the frequency offset is 3.5 ppm,
or 224 Hz at 1.5 T. The degree of water–fat shift
(WFS) is inversely proportional to the frequency-
encoding bandwidth, as with susceptibility. The
prevalence of type 1 artifact to the frequency-
encode direction (for non-EPI sequences) is due
to the relatively low readout BW of these
sequences, as well as the fact that transverse
magnetization is either refocused or spoiled for
each echo measurement. This effectively negates
the accumulation of precession-related offsets in
the phase-encode direction. Since EPI utilizes a
very high BW, it is a special case where type 1
artifact typically occurs in the phase-encode
direction.
4.5.4 Resolution
Type 1 artifact does not degrade image quality to the
extent of motion or susceptibility artifact. However,
it may have impact on diagnosis if water–fat misre-
gistration obscures pathology such as cartilage thick-
ness. As alluded, chemical shift is greatest when the
frequency-encode BW is low. But it also becomes
more significant when image resolution is also low
(large FOV or small matrix). Since the resonant offset
between water and fat is well defined (224 Hz at
Fig. 4.14 (a, b) Axial T1 turbo spin
echo of the soft tissues of the neck
reveal subtle black etching around
tissue boundaries (arrows).
69
| 09.11.19 - 01:53
Magnetic Resonance Imaging
1.5 T and 448 Hz at 3 T, etc.), one can easily calculate
the expected WFS in terms of image pixel offset, if
the FOV, matrix, and BW are known (see appendix).
Since FOV and matrix are often fixed due to applica-
tion criteria, increasing BW directly remedies the
artifact (▶ Fig. 4.15). This tactic also has less SNR
penalty than increases in image resolution, which is
another remedy. Alternatively, frequency and phase
directions can be swapped, in lieu of any parameter
adjustment, as long as other artifacts, such as
70
Fig. 4.15 Axial T1 turbo spin echo of
the neck at two different bandwidths:
200 Hz/px (left), and 425 Hz/px (right).
Note that the frequency-encoding
direction is left-right in (a) and ante-
rior-posterior in (b). There is noticeable
reduction in chemical shift signal loss
along boundaries with higher
bandwidth.
aliasing and motion, are not adversely affected. One
must also be wary of chemical shift at high field
strengths, since one-to-one transfer of imaging
parameters will not be optimal; a proportional
increase in BW is necessary to achieve the same
WFS as lower field strength. Another solution to
eliminate the appearance of fat shifts is to employ fat
saturation. Though effective, this clearly alters the
purpose of the sequence, and may not be warranted
in particular clinical applications.
| 09.11.19 - 01:53
4.6 Case 6: Application of Fat-Suppressed Sequences
4.6 Case 6: Application of Fat-
Suppressed Sequences
in the Pelvis did not
Reveal the Expected Contrast
4.6.1 Background
● Fat-suppressed T1 and T2 sequences are univer-
sally employed in body MR applications, such as
female pelvic imaging for fibroids.
● The most common fat saturation technique is
chemically selective RF excitation, whereby only
fat resonant frequency is specifically targeted
with a saturation pulse.
4.6.2 Findings
● Chemical fat saturation is incomplete in both
T1- and T2-weighted imaging (▶ Fig. 4.16).
● It is likely that the MR system erroneously tuned
the center frequency to be the dominant species
in the FOV, in this case fat.
● Alternatively, strong inhomogeneous fields or
poor shimming within the FOV may alter the
expected off-resonance position of fat species
relative to water.
● Chemical fat saturation RF pulses are automati-
cally applied at 440 Hz (at 3T) down field from
the tuned center frequency, which no longer
effectively excite fat resonances.
● Partial or incomplete fat saturation typically
appears as regional bands of dark fat tissue.
Fig. 4.16 Fat-suppressed axial (a) T2 single-shot, and (b) T1 3D gradient echo acquisitions did not produce hypointense
fat signal as expected. Some regional fat uppression is evident.
4.6.3 Discussion
Fat suppression has many applications, including
making edema and inflammation more conspic-
uous on T2-weighted imaging and eliminating
confounding high signal from postcontrast
T1-weighted imaging. One important advantage of
a resonant frequency difference between water
and fat protons is the ability to selectively saturate
the magnetization of fat in MR images. This can be
done in a variety of ways, but the most common is
to center an additional RF saturation pulse over
the resonant frequency of fat. In actuality, fat has
up to six different resonant frequencies, with the
most significant occurring at 1.3, 2.1, and 0.9 ppm
(▶ Fig. 4.17). For this reason, spectrally selective RF
pulses must also have a prescribed BW, but must
be limited to prevent intruding water resonance at
4.7 ppm. In some areas of the body where complex
or abnormal tissue geometry affects the local
magnetic field, the prescribed fat-centered RF
pulse and BW may partially “miss” the true
susceptibility-induced resonant frequency of fat
causing incomplete tissue saturation on images.
Moreover, this scenario may even lead to erro-
neous saturation of water signal (▶ Fig. 4.18a).
Another root cause for poor or erroneous fat
saturation is the inefficiency of RF pulses them-
selves; sharp frequency cutoffs are difficult to
achieve, especially over a small spectral range.
Consequently, the bell-shaped profile may cause
some varying excitation of resonances inside and
outside the frequency bounds.
Separate from local susceptibility changes, large
FOV imaging also causes regions of poor fat sup-
pression, primarily along the periphery of the FOV.
71
| 09.11.19 - 01:53
Magnetic Resonance Imaging
Fat-containing voxels located along the periphery
are significantly far from isocenter, where field
inhomogeneity also predominates. This also
pertains to multislice axial imaging; poor fat sup-
pression is often seen on first and last slices of
axial data sets with large number of slices.
4.6.4 Resolution
An immediate solution to poor fat suppression is
improving the fat suppression pulses themselves.
Using adiabatic RF excitation helps improve the
spectrum of targeted fat protons. Alternatively,
spectral excitation can be performed on water
protons only, whose spectral amplitude and line
width are usually more well defined than fat.
Another strategy is to convert the acquisition to a
short-tau inversion recovery (STIR) technique
which offers increased suppression uniformity
over broad FOVs and field inhomogeneity
(▶ Fig. 4.18b). STIR utilizes a nonselective 180-
degree inversion (IR) prepulse timed to null the
72
longitudinal magnetization of fat protons prior to
image acquisition, based on the its short T1 value
(~ 250 millisecond at 1.5T). Since the IR prepulse
affects both fat and water proton resonant
frequencies, all tissue will undergo longitudinal
T1 recovery. Most tissues relax slower than fat,
and will not be suppressed at the selected inver-
sion time (TI); however, they will incur reduced
available magnetization, which translates to
reduced image SNR.
With chemically selective fat suppression
methods, it is always useful to observe the spectral
peaks of fat and water following any shimming
procedure when fat suppression uniformity is
desired. Even though broad line widths may still
persist, manual frequency adjustments help to
resolve significant fat frequency shifts caused by
off-resonance. This strategy can be further opti-
mized by using smaller FOVs, or fewer slices,
thereby limiting the effective volume of shimming.
More systems are now equipped with sophisti-
cated fat-water separation techniques, which
Fig. 4.17 Sample spectrum of fat and
water resonant peaks from a voxel
obtained in fatty liver tissue. Note the
broad lipid peak, suggesting the pres-
ence of other lipid resonances. Arbitrary
units are used along the x-axis.
Fig. 4.18 Comparison of (a) spectrally
selective fat suppression, and
(b) nonselective inversion-recovery-
based fat suppression (STIR) in the
cervical spine. Local field homogeneity
impacts the uniformity of fat suppres-
sion in (a) to the extent that water
signal may also be partially affected.
| 09.11.19 - 01:53
4.6 Case 6: Application of Fat-Suppressed Sequences
evolved from the well-known two-point Dixon
method (▶ Fig. 4.19). Modern versions of the
method still incur a scan time penalty, but are very
efficient for creating robust fat-suppressed (water-
only) images. These techniques do not rely on
spectrally selective pulses, or nonselective IR
Fig. 4.19 Comparison of (a) spectrally
selective fat suppression, and (b) two-
point Dixon water-only images of the
neck in two different subjects. While
some expected inhomogeneous is
expected in (a), more robust suppres-
sion is achieved by separating fat and
water images with the Dixon tech-
nique (b).
pulses, which reduce SNR of all tissues. However,
the overall efficacy of the reconstruction is highly
dependent on producing a suitable domain
magnetic field map. Nonetheless, unfavorable fat/
water swapping can persist in regions of signifi-
cant Bo field inhomogeneity.
73
| 09.11.19 - 01:53

Magnetic Resonance Imaging

4.7 Case 7: T1-Weighted 4.7.3 Discussion

Gradient Echo of the Abdomen The availability of parallel imaging methods has
tremendously improved the utility of MR in a
Shows Marked Artifact Medially variety of applications. Parallel imaging involves
on Both Coronal and Axial FOV, utilizing multiarray receive coils over the
Obscuring Visualization of Soft imaged region, and using their individually
specific coil sensitivity to reconstruct under-
Tissues sampled k-space data. Scan times can be drasti-
cally reduced by 2 × or more with parallel
4.7.1 Background imaging but the side effect is reduced SNR and
● Most T1-weighted 2D and 3D acquisitions of the some associated artifacts. Often significant
chest, abdomen, and/or pelvis require subject amplification of noise is seen with parallel
imaging when reduction factors exceed 3 × .
breath hold and some method of scan acceleration.
Furthermore, if multiarray coils are not properly
● In addition to reducing the phase-encode resolu-
placed around the region of interest, more noise
tion, utilizing parallel imaging acceleration with amplification is observed. One parallel imaging
multichannel receive coils is another common method reconstructs undersampled k-space
strategy. data using multiple coil sensitivity images in
image space (e.g., SENSE). Artifacts with this
4.7.2 Findings method appear similar to image foldover,
although the aliased regions typically appear in
● Both examples in ▶ Fig. 4.20 utilize parallel the center of the FOV and are sometimes
imaging approaches. mistaken as signal “hotspots.” Another common
● Relatively small FOV was prescribed in the parallel imaging technique estimates missing
imaging information in k-space (e.g., GRAPPA).
respective phase-encode directions to further
If missing k-space data is not effectively recov-
reduce acquisition time.
ered, unwanted phase shifts may develop,
● Both images suffer from unwanted signal artifact resulting in ghost-like artifacts in the phase-
in relevant soft tissue. encode direction (▶ Fig. 4.21).

Fig. 4.20 (a) T1-weighted 3D gradient echo (GRE) of the abdomen shows marked artifact through the middle
of the field of view, obscuring visualization of soft tissues. (b) A nonspecific high-signal line also appears on
an axial 2D GRE.

74
| 09.11.19 - 01:53
4.7.4 Resolution
Parallel imaging artifacts may mimic physiological
or other technical artifacts; therefore, one must
first identify the true origin of poor SNR, ghosts, or
aliasing. For image-based parallel imaging, some
extended phase FOV (oversampling) will alleviate
the subtle unfolding reconstruction artifact.
Furthermore, effort should be made to match the
anatomic positioning between coil calibration
scans and pulse sequences using parallel imaging.
This may require calibration scans to be performed
using similar breath hold instructions.
4.7 Case 7
Fig. 4.21 Subtle ghost-like artifact
appears in liver across the center of the
field of view (arrows). This artifact is
not indicative of motion, but rather
results from parallel imaging effects.
In k-space-based methods, parallel imaging
ghosts are usually rare since autocalibration
steps are built into the acquisition. However, the
technique is more sensitive to patient motion or
inadequate reference coil sensitivities. Often,
more autocalibration reference lines are needed,
which reduces the scan acceleration. Increased
reference lines also alleviate central noise
banding, which is common with parallel imaging.
Finally, routine system and coil maintenance is
vital for optimal performance of sequences using
parallel imaging.
75
| 09.11.19 - 01:53

Magnetic Resonance Imaging

4.8 Case 8: Abnormal Dark 4.8.2 Findings

Fluid Seen in the Bladder ● While each slice in HASTE imaging is acquired
of an Axial Single-Shot T2- individually in less than a half second, an entire

Weighted Sequence, but volume of tissue is excited and combined every

not on Location-Matched
3D T2 Acquisition
4.8.1 Background
● All soft-tissue pelvic imaging involve either large
FOV axial T2 for gross anatomy, or small FOV
(high resolution) T2 for evaluating specific
diseases, such as rectal and prostate cancer.
● Often, half-Fourier-acquired single-shot turbo
spin echo (HASTE) is performed for overall
anatomical assessment since it is able to quickly
acquire large superior-to-inferior coverage with
minimal motion artifacts.
● HASTE acquires just over half of its k-space data
in one TR period, with the rest being interpo-
lated. This enables ultra-fast single-slice
acquisition, without T1-weighting.
● High resolution 3D T2 acquires k-space data in
multiple TR periods, depending on the amount
of total phase-encode data and the acquisition
echo train length.
TR period for the 3D T2 technique.
● Static fluid in the bladder will be T2-bright by
definition on both sequences, but some focused
regions of hypointense signal indicate some flow
on HASTE (▶ Fig. 4.22a).
● The ultrafast acquisition of HASTE will be
sensitive to some flow turbulence entering the
slice during the acquisition. This fluid was not
originally excited, but replaces some fluid in this
course of time.
● 3D T2 is not sensitive to this subtle fluid motion
since its influence is effectively averaged out
over the multiple TR periods.
4.8.3 Discussion
The case above exemplifies the ever-present
concept of flow-related contrast in MRI. In many
cases, such as in angiographic applications, flow-
related contrast is expected with sequences espe-
cially configured to be sensitive to the phenomena.
However, many essential sequences intended for
soft tissue analysis may be prone to unexpected

Fig. 4.22 Abnormal dark fluid seen in the bladder of an axial single-shot T2-weighted sequence (a). The appearance is
absent from a location-matched 3D T2 acquisition (b).

76
| 09.11.19 - 01:53

4.8 Case 8: Abnormal Dark Fluid Seen in the Bladder

hyper- or hypointensities related to flow. It was
previously discussed (Case 1) that fluctuating signal
intensity due to periodic flow will lead to ghosting
if not properly compensated or synchronized with
the acquisition. While this is particularly true for
TSE sequences, flow-related contrast may also
appear as a static indication of a particular flow
state and possibly even point to disease abnormal-
ities. In the case of ▶ Fig. 4.22a, the fluid in the
bladder is mostly stationary, and predictably
appears bright due to its high T2 value. The dark
areas represent subtle flow from fluid entering the
slice between excitation and TE. In contrast,
consider ▶ Fig. 4.23a, b, where blood in the aorta
and portal vein are both dark on axial and coronal
HASTE, respectively. In rapid flowing vessels, blood
that is slice-excited quickly travels out of the slice
before the signal is both refocused and measured
(at TE), leaving a “flow void.” As flow becomes
restricted or stationary (as in veins or certain parts
of the cardiac cycle), blood will appear diffusely
hyperintense, similar to fluid, since it also has rela-
tively high T2. This is apparent in ▶ Fig. 4.23c, d, in
a patient with metastatic neuroendocrine disease,
where HASTE shows some hyperintense blood
signal representative of slow or restricted flow. In
this case, it is important to examine all slices to
determine whether the signal is due to sensitivity
to the cardiac cycle or related to disease (such as
hypertension or thrombosis). Flow-related contrast
also affects gradient echo MR sequences. Since TE
and TR are typically shorter than spin echo type
sequences, unperturbed blood flowing into the slice
is both excited and measured in close tandem,
producing a bright blood phenomenon. A well-
known example is 3D time-of-flight (TOF) imaging.

Fig. 4.23 (a) Axial and (b) coronal T2 single-shot in a healthy subject reveals hypointense signal in the aorta and portal
vein. Similar acquisitions obtained in a patient with metastatic disease (c and d) show some hyperintense blood signal in
the corresponding vessels, suggesting slow or restricted flow.

77
| 09.11.19 - 01:53
Magnetic Resonance Imaging
Fig. 4.24 Examples of entry-flow phenomena on gradient echo (GRE) images. (a) 2D axial GRE, with bright blood in liver
vessels; (b) a reconstructed coronal-oblique 3D GRE with hyperintense blood signal, which slowly fades in the direction
of flow.
In this sequence, blood entering the imaging
volume is excited and measured at TE, but escapes
the volume before the next excitation. This “entry-
flow” enhancement effect can be contrasted with
stationary tissue that remains dark due to repeated
exposure to RF excitation. In some cases, bright
blood on gradient echo is unwanted or unexpected,
as shown in ▶ Fig. 4.24, since it may be mistaken for
the presence of gadolinium contrast agent.
4.8.4 Resolution
As stated, flow-related contrast is desirable in
many applications. In addition, it aids in many
diagnostic instances. In practice, it is important to
understand how flow from generally static fluid or
from fat flowing blood may affect the signal prop-
erties of specific spin or gradient echo sequences.
78
Lack of flow voids on spin echo type imaging,
especially HASTE, are indicative of slow-moving
blood remaining in the imaging slice. These
diastolic hyperintensities can be lessened with
cardiac gating, with or without the addition of a
“black blood” pulse. These additional pulses come
in a variety of methods (both gradient and RF
forms), and help to further eliminate residual
blood signal though flow cancellation and
dephasing. Conversely, flow voids are unwanted in
angiographic 3D gradient echo. It is important
to prescribe sufficient volumetric coverage to
maximize inflow enhancement, while being
conscious that in-plane flow from tortuous vessels
may result in some signal loss. For nonangio-
graphic gradient echo, shortened TR and opti-
mized TE, especially for 3D imaging, ensures dark
vessel lumen.
| 09.11.19 - 01:53
4.9 Case 9: Postcontrast T1-Weighted Gradient Echo Reveals Patchy Enhancement
4.9 Case 9: Postcontrast T1-
Weighted Gradient Echo
Reveals Patchy Enhancement in
the Anterior Septal Wall
4.9.1 Background
● Corresponding cine and postcontrast T1 is
essential for demonstrating areas of scar due
to myocardial infarction.
● Postcontrast T1 is commonly performed with
inversion recovery fast gradient echo sequences,
which is intended to enhance the contrast
between healthy and infarcted myocardium.
● T1 image contrast is controlled by the time
delay between the inversion pulse and data
acquisition.
4.9.2 Findings
● Area of scar enhancement not clearly defined
relative to healthy myocardium.
● Corresponding short-axis cine shows thin
myocardium in the same region.
4.9.3 Discussion
Delayed contrast enhancement (DCE) is a
powerful cardiac MRI technique for distinguishing
areas in the myocardium with increased extracel-
lular distribution volume due to cell necrosis,
fibrosis, or scarring (▶ Fig. 4.25). This extends
Fig. 4.25 (a) Postcontrast T1-weighted gradient echo reveals patchy enhancement in the anterior septal wall.
(b) Corresponding short-axis cine shows thin myocardium in the same region.
from the pharmacokinetics of most gadolinium
contrast agents, which provide significant T1 en-
hancement in regions of high vascularity and
extracellular space. In the case of infarcted
myocardium, the scarred region experiences
delayed contrast distribution due to poor vascu-
larity followed by prolonged clearance. This is in
stark contradiction to the more predictable
kinetic behavior in both remote myocardium and
blood pool. While fast T1-weighted imaging can
exploit the T1 differences between normal and
infarcted myocardium, image contrast can be
significantly improved by implementing IR prepa-
ration pulses. These 180-degree RF pulses, which
invert the polarity of all available longitudinal
magnetization, are performed at a designated
time prior to normal data acquisition. In other
words, this process “prepares” the longitudinal
magnetization for subsequent data collection.
Since all tissues including normal and infarcted
myocardium have unique T1 values, IR pulses
have the ability to further enhance the dynamic
range and contrast if the time of inversion (TI) is
selected appropriately. Maximum apparent
contrast in the case of delayed myocardial
imaging is achieved when TI is selected to cancel
(or “null”) the longitudinal recovery of normal
myocardium (▶ Fig. 4.26). The TI null point
implies tissue signal suppression (in terms of its
T1 value), since no longitudinal magnetization
exists as imaging data collection commences.
▶ Fig. 4.27 shows another example of how
changing the TI value significantly alters the
ability to distinguish infarcted from normal
myocardium.
79
| 09.11.19 - 01:53

Magnetic Resonance Imaging

Fig. 4.26 Plot of the relative longitudinal T1 recovery of contrast-enhanced normal and infarcted myocardium following
application of an inversion (180 degrees) RF pulse. The concept is to commence image acquisition after an inversion
time (TI) set to suppress the signal of normal myocardium. This occurs at the point when recovery crosses the x-axis.

Fig. 4.27 Comparison of contrast-enhanced short-axis inversion recovery in a subject with subendocardial infarction.
A TI of 600 milliseconds (a) does not reveal the extent of infarct region, while adapting to TI = 300 milliseconds (b) shows
the enhancement sufficiently (arrow) by virtue of suppressing signal from normal myocardium.

80
| 09.11.19 - 01:53
4.9 Case 9: Postcontrast T1-Weighted Gradient Echo Reveals Patchy Enhancement
There are a host of applications that utilize IR.
Each carefully prescribes TI to null a particular
tissue’s (known) T1, or enhance contrast between
two T1 s. Both STIR- and fluid-attenuated inversion
recovery (FLAIR)-type sequences employ IR to null
fat and fluid, respectively. Some other applications,
such as postcontrast brain imaging (e.g., MPRAGE),
utilize an optimized TI to enhance grey and white
matter differences, in contrast to enhancing
lesions. It can be shown that T1 image contrast
improves with the use of IR prepulses at the
expense of some increased imaging time.
4.9.4 Resolution
In general, if T1 is known (e.g., T1fat = 250 milli-
second; T1csf = 3500 millisecond), optimal null
point TI is calculated from a modified monoexpo-
nential recovery equation (TI = 0.693 × T1). An
analytical solution can also be made to estimate
the expected image contrast of any chosen TI,
given multiple T1 values
S ¼ 1  2eTI=T1
▶ Fig. 4.27 shows the important image contrast
differences when different TIs are chosen for the
same application. In this case, image contrast
between subendocardial infarct and normal
myocardium is optimized with TI = 250 milli-
second (▶ Fig. 4.27b) compared to TI = 600 milli-
second (▶ Fig. 4.27a). However, since contrast
agent concentration changes dynamically
following administration, simply applying TI = 250
millisecond to the case in ▶ Fig. 4.25 may not
always null normal myocardium. Often, a “TI-
scout” acquisition must be performed, which
quickly samples images at multiple TI values. The
user then selects the optimal null point TI to
subsequently optimize IR delayed enhancement
imaging. Similar TI optimization is needed in other
sequences, as well, such as FLAIR. In this sequence,
CSF fluid suppression is desired. Since CSF fluid
has a lengthy T1 (~ 3500 millisecond), which is on
the order (or greater than) most spin echo TRs, one
must consider the effect of incomplete T1 relaxa-
tion when choosing an optimal TI value. In this
case, the optimal TI is less than the predicted TI,
and scales exponentially with TR. For complete-
ness, it is often important to consider other
parameters, such as TR, flip angle, and echo train
length, when determining optimal TI, especially
for long T1 tissues.
81
| 09.11.19 - 01:53

Magnetic Resonance Imaging

4.10 Case 10: Significant integrity in any part of these two elements
dramatically reduces the upper threshold of
Signal-to-Noise Variation achievable SNR and, hence, image quality. For
Across the FOV, this reason, it is important to have a routine
quality assurance program in place, whereby RF
Creating Nondiagnostic coils are tested for proper functionality. An
Image Quality important distinction must be made when inter-
rogating errors and artifacts related to either RF
4.10.1 Background transmission or RF reception. Particularly, RF
transmission field mostly originates from the
● MR signal originates from net magnetization
main magnetic coil located within the bore (even
aligned along the main magnetic field.
though local transmit/receive and dual transmit
● Above all, the imaging process requires spin coils also exist). This transmission is pretuned
excitation and signal reception by RF coils. based on desired flip angles and emits across a
● The homogeneity of the prescribed transmission broad volume within the bore. It is clear that the
and receiving RF fields are proportional to the integrity of the RF transmission field is only
entirely uniform over a smaller finite volume
resulting image SNR.
(~ 30–40 cm) around the isocenter, and falls off
thereafter due to inhomogeneity (▶ Fig. 4.30).
4.10.2 Findings The result is variation in the expected flip angle
distribution, which in turn affects transverse
● Significant noise is apparent in the middle of
magnetization. Another dramatic effect of RF
FOV, and anterior to the spinal anatomy transmit field inhomogeneity is dielectric effect
(▶ Fig. 4.28). (▶ Fig. 4.31). In this situation, which mostly
● Underlying T2 image contrast is sufficient, occurs at high field strengths (> 1.5T), image
despite poor SNR. signal variation is caused by the RF field’s inter-
action with various tissue conductivities, particu-
● Brain anatomy is spared from poor image quality.
larly increased water content.
RF receive coil integrity is equally (if not more)
4.10.3 Discussion important for optimal SNR MRI. It is entirely
possible (and common) that a well-functioning
Upon closer inspection, one finds that a partic- receive coil can produce low SNR images, if care
ular receive coil element surrounding the neck
was mistakenly deactivated during the scan.
Fortunately, recognizing this error allows a
simple remedy to ensure all necessary coils are
activated over the region of interest, thus recov-
ering the underlying signal (▶ Fig. 4.29). The
persistently lower SNR anterior to spinal
anatomy points to the use of primarily posterior
receiver coil elements. While this scenario and
remedy seems trivial, other more significant root
causes may lead to SNR variation across the FOV,
such as malfunctioning receiver coils. The
common theme, therefore, is to recognize the
unequivocal importance of both receive and
transmission RF fields in MR image quality. As
stated in the background, these RF fields buttress
the entire MR image formation process; RF trans-
mission begins the experiment by exciting
proton spins, while RF receiver coils (or Fig. 4.28 Significant signal-to-noise variation across the
“antennas”) capture the encoded signal. Loss of field of view, creating nondiagnostic image quality.

82
| 09.11.19 - 01:53

4.10 Case 10: Significant Signal-to-Noise Variation Across the FOV,

Fig. 4.29 Comparison of the original

acquisition (a), with a second acquis-
ition with the appropriate coils
activated (b).

Fig. 4.30 (a) An initial localizer

acquisition uses a large field of view
(FOV) to routinely locate and identify
the anatomy in question. It also reveals
the effective region coil sensitivity and
signal fall-off. (b) A smaller, more
focused FOV is subsequently used for
higher-quality anatomic visualization,
while being conscious to limit
including areas outside the effective
sensitivity region.

Fig. 4.31 (a) The increased abdominal

fluid affects the RF transmit
uniformity. The fluid acts as dampener
due to the change transmit conduc-
tivity, thereby modifying the effective
flip angle. (b) The dielectric effect is
absent in patients with minimal to no
interabdominal fluid.

is not paid to its proper positioning. This is actuality, the coil was placed suboptimally. The
exemplified in ▶ Fig. 4.32, where there may be skill developed in coil use and placement is often
legitimate question about the significant drop in understated and presumed.
SNR across a multislice acquisition, when in

83
| 09.11.19 - 01:53

Magnetic Resonance Imaging

Fig. 4.32 (a) Sagittal T2 image of the pelvis does not extend through the pelvic floor due to poor coil coverage in
the region. (b) and (c) show corresponding axial acquisitions from two locations. Note the significant drop in SNR
in (c) relative to (b) toward the pelvic floor due to poor coil sensitivity.

4.10.4 Resolution FOV < 30 cm to preserve optimal quality. Initial

Many image quality issues stem from poor coil
placement or not recognizing the optimum extent
of the RF transmit uniformity. Quality MR would
not be possible without careful attention to both
items. If an imaging practice intends to perform
large FOV exams (e.g., abdomen-pelvis, thoracic
spine, long bones, etc.), it is wise to acquire large
(> 40 cm FOV) survey acquisitions, and measure
the region of optimal signal homogeneity avail-
able. This can be appreciated in ▶ Fig. 4.31, where
a FOV of 40 cm shows signal drop out beyond a
30 cm radius; subsequent scans are limited to
survey scans also reveal whether receive coils are
positioned too high or low (▶ Fig. 4.32). This
should be adjusted based on the extent of
anatomic coverage. Positioning and activation of
coils and coil elements should also play a role
when diagnosing other artifacts, such as parallel
imaging (Case 7, T1-Weighted Gradient Echo
(GRE) of the Abdomen Shows Marked Artifact
Medially on Both Coronal and Axial FOV,
Obscuring Visualization of Soft Tissues) and
aliasing (Case 3, Appearance of Extra Field-of-
View Anatomy on the Inferior Portion of Sagittal
3D T2-Weighted Acquisition of the Spine).

84
| 09.11.19 - 01:53
4.11 Review Questions
4.11.1 Case 1: Appearance of
Discrete Image Ghosts on
Abdominal Imaging
1. The acquisition of phase-encode steps at both
inspiration and expiration states during the
course of free-breathing abdominal MRI will
most likely result in what appearance in the
image?
a) Image blurring
b) Discrete image ghosts
c) Ringing in the frequency direction
d) Pulsation artifacts
2. Which of the following methods is NOT an
effective strategy for reducing motion
artifacts?
a) Breath holding
b) Respiratory gating
c) Shallow breathing
d) Ultrashort TR sequences
3. “Single-shot” MR abdominal acquisitions, such
as half-Fourier turbo spin echo (TSE), collect all
phase-encode steps from one slice in one TR
period, with very short echo spacing. Regarding
motion artifacts due to free breathing, which of
the following is NOT true about single-shot
methods?
a) Multislice images may be misregistered with
one another
b) Subtle ghosts will be apparent along the
phase-encode direction
c) Image blurring may occur
d) Cross-talk and signal saturation may occur
between sequentially acquired slices
4.11 Review Questions
4.11.2 Case 2: A Well-Defined
Area of Signal Hyperintensity
Appears Bilaterally at the Level
of the Internal Auditory Canal on
Diffusion-Weighted MRI, Affecting
Visualization of Surrounding
Structures
4. Which of the following strategies will help
reduce metal susceptibility artifact?
a) Increase BW, and switch to fast gradient
echo
b) Increase BW, and lower TE
c) Switch to fast gradient echo, and increase
resolution and signal averaging
d) Swap phase-encode direction, and increase
resolution and slice thickness
4.11.3 Case 3: Appearance of
Extra Field-of-view Anatomy on
the Inferior Portion of Sagittal 3D
T2-Weighted Acquisition of the
Spine
5. To image small anatomy, such as the pituitary,
without aliasing artifact using an FOV smaller
than the brain, which of the following tactics
should NOT be employed?
a) Sagittal 2D imaging with maximum phase
oversampling
b) Axial 3D imaging with nonselective RF
pulses and maximum phase oversampling
c) Coronal slab-selective 3D imaging some slice
oversampling
d) Axial 2D imaging using a rectangular FOV in
the left-right direction, such that the ears
are truncated
85
| 09.11.19 - 01:53
Magnetic Resonance Imaging
4.11.4 Case 4: Precontrast, Axial
3D T1-Weighted Gradient Echo
with Fat Suppression Shows
Adequate Anatomical Detail, but
Minor Edge Ripple and Blur that is
Presumed to be Motion
6. In 3D imaging, how can one better distinguish
truncation from motion artifact?
a) Reconstruct data in another dimension
b) If ringing occurs in all directions, it is due to
motion
c) Signal intensity is much higher with
truncation
d) There is no way to distinguish them
7. Truncation artifact may occur in the frequency-
encode direction when:
a) Resolution is very high
b) BW is large
c) There are very high-contrast boundaries
along that direction
d) A smoothing filter is applied
4.11.5 Case 5: Dark Etching
Appears at the Boundary of Fat
and Soft-Tissue Layers
8. A water-fat shift (WFS) of two pixels at 1.5 T
implies:
a) Water and fat are off-resonant by 2 times
224 Hz (448 Hz)
b) Water and fat will be misregistered by two
pixels in the phase-encode direction
c) The BW in Hz/pixel must be half of 224 Hz
(112 Hz)
d) FOV must be too large
9. Which imaging scenario would reduce the
conspicuity of type 1 chemical shift?
a) Removing fat sat from EPI
b) Swapping phase- and frequency-encode
directions
c) Switching from spin echo to gradient echo
sequences
d) Increasing the imaging resolution
86
4.11.6 Case 6: Application of
Fat-Suppressed Sequences in the
Pelvis did not Reveal the Expected
Contrast
10. All of the following strategies will help to
improve fat suppression, except:
a) Repositioning the subject at the magnet
isocenter
b) Switching to STIR acquisitions
c) Switching receiver coils
d) Shimming and manually adjusting the
center frequency
11. Frequency-selective fat saturation is usually
most uniform in which of the following
scenarios?
a) Anatomy that allows small FOV, such as the
knee
b) Anatomy with irregular geometry, such as
the neck
c) Off-center FOV, such as the elbow at one’s
side
d) Large FOV, such as bilateral femurs and hips
4.11.7 Case 7: T1-Weighted
Gradient Echo of the Abdomen
Shows Marked Artifact Medially
on Both Coronal and Axial FOV,
Obscuring Visualization of Soft
Tissues
12. Given that SNR decreases with the use of parallel
imaging, what can be done to improve relative
SNR and lessen parallel imaging artifacts?
a) Oversample in the phase-encode direction
b) Reduce resolution
c) Increase slice thickness
d) Use more slices
13. In spine MRI, typically only posterior receive
coils are used. In terms of using parallel
imaging, this implies:
a) Parallel imaging can be used in any direction
b) For sagittal imaging, it is best to use parallel
imaging in the superior-inferior direction
| 09.11.19 - 01:53

4.11 Review Questions

c) For axial imaging, parallel imaging in the b) SNR is too low

anterior-posterior direction is preferred c) The short TI may inadvertently suppress
d) Parallel imaging should not be used contrast agent effects
d) Too many flow artifacts

4.11.8 Case 8: Abnormal Dark 17. The null point TI is modified from 225 to 300
Fluid Seen in the Bladder of an millisecond during two DCE scans 5 minutes
Axial Single-Shot T2-Weighted apart. This implies:
Sequence, but not on Location- a) Contrast agent is mostly in the blood pool
Matched 3D T2 Acquisition b) Fat tissue needs to be suppressed for optimal
contrast
14. Low signal intensity of cerebrospinal fluid
c) TR is also short
(CSF) on T2-weighted cervical spine imaging is
d) Contrast agent is slowly being cleared from
likely due to:
normal myocardium
a) Flow ghosting
b) Use of saturation bands
c) Spin dephasing due to pulsatile flow
4.11.10 Case 10: Significant
d) Flow compensation Signal-to-Noise Variation Across
the FOV, Creating Nondiagnostic
15. T2-weighted TSE sequences are more sensitive Image Quality
to flow voids than T1-weighted TSE because: 18. Another possible cause for signal loss in the
a) T2 values of blood are short following image is:
b) TE is higher, giving more time for flowing
protons to “escape” the imaging slice
c) The multitude of 180 refocusing pulses
associated with T2 TSE continually saturates
flow
d) T1 TSE sequences allow more complete
recovery of flow signal each TR period

4.11.9 Case 9: Postcontrast T1-

Weighted Gradient Echo Reveals
Patchy Enhancement in the
Anterior Septal Wall
16. STIR acquisitions are typically not used for
postcontrast imaging because:
a) Spectrally selective fat saturation is good
enough

87
| 09.11.19 - 01:53

Magnetic Resonance Imaging

a) Cardiac motion and flow Glossary

b) Metal artifact
B0: main magnetic field strength
c) A long MR bore
TR: repetition time
d) Having too many coil elements activated
TE: time to echo
Ny: pixels in the phase (y) direction
19. One remedy for dielectric effect seen in the
Nx: pixels in the frequency (x) direction
following image is:
Nz: pixels in the slice (z) direction
NSA: number of signal averages
Δx: resolution (mm) in the frequency direction
Δy: resolution (mm) in the phase direction
Δz: resolution (or slice thickness) in the slice
direction
BWread: bandwidth in the frequency readout
direction
WFS: water–fat shift (in pixels)
FOV: field-of-view
Δf: relative frequency offset between two proton
types (e.g., fat and water)
Δχ: susceptibility difference between two tissues
g: geometric factor related to coil configuration
and placement
R: parallel imaging acceleration factor

a) Position the patient prone Suggested Reading

b) Omit T2 imaging from the protocol Bernstein MA. Signal acquisition and k-space
c) Reposition the receive coil sampling. In: Bernstein MA, 1st ed. Handbook of
d) Attempt the exam at lower field strength, MRI Pulse Sequences. Burlington, US: Elsevier
if possible Academic Press; 2004:367–442
Brown MA, Semelka RC. MRI: Basic Principles and
Application. 2nd ed. New York, NY: John Wiley &
Equations Sons, Inc; 1999
Brown RW, Cheng YC, Haacke EM, Thompson MR,
Scan time ¼ TR  Ny  Nz  NSA
Venkatesan R. Magnetic Resonance Imaging: Phys-
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! ical Principles and Sequence Design. 2nd ed.
NSA  Nx  Ny  Nz Hoboken: Wiley; 2014:944
SNR / K ðΔx  Δy  ΔzÞ
BWread
Graves MJ, Mitchell DG. Body MRI artifacts in clin-
ical practice: a physicist’s and radiologist’s
FOV  Δf
WFS ¼ perspective. J Magn Reson Imaging. 2013; 38(2):
Nx  BW
269–287
Susceptibility artifact size e ðΔxÞ  TE  B0 =BWread Griswold MA, Jakob PM, Heidemann RM, et al.
Generalized autocalibrating partially parallel
1 acquisitions (GRAPPA). Magn Reson Med. 2002; 47
SNR e pffiffiffi
g R (6):1202–1210

88
| 09.11.19 - 01:53
Hargreaves BA, Worters PW, Pauly KB, Pauly JM,
Koch KM, Gold GE. Metal-induced artifacts in MRI.
AJR Am J Roentgenol. 2011; 197(3):547–555
Hamilton J, Franson D, Seiberlich N. Recent
advances in parallel imaging for MRI. Prog Nucl
Magn Reson Spectrosc. 2017; 101:71–95
Huang SY, Seethamraju RT, Patel P, Hahn PF, Kirsch
JE, Guimaraes AR. Body MR imaging: artifacts, k-
space, and solutions. Radiographics. 2015; 35(5):
1439–1460
Lee VS. Cardiovascular MRI: Physical Principles to
Practical Protocols. Philadelphia: Lippincott
Williams & Wilkins, 2006
Mugler JP, III. Optimized three-dimensional fast-
spin-echo MRI. J Magn Reson Imaging. 2014; 39
(4):745–767
Pruessmann KP, Weiger M, Scheidegger MB,
Boesiger P. SENSE: sensitivity encoding for fast
MRI. Magn Reson Med. 1999; 42(5):952–962
Qiu D, Chan GC, Chu J, et al. MR quantitative
susceptibility imaging for the evaluation of iron
4.11 Review Questions
loading in the brains of patients with β-thalas-
semia major. AJNR Am J Neuroradiol. 2014; 35(6):
1085–1090
Reeder SB, Atalar E, Bolster BD, Jr, McVeigh ER.
Quantification and reduction of ghosting artifacts
in interleaved echo-planar imaging. Magn Reson
Med. 1997; 38(3):429–439
Simonetti OP, Kim RJ, Fieno DS, et al. An improved
MR imaging technique for the visualization of
myocardial infarction. Radiology. 2001; 218(1):
215
Wang J, He L, Zheng H, Lu ZL. Optimizing the
magnetization-prepared rapid gradient-echo (MP-
RAGE) sequence. PLoS One. 2014; 9(5):e96899
Wheaton AJ, Miyazaki M. Non-contrast enhanced
MR angiography: physical principles. J Magn Reson
Imaging. 2012; 36(2):286–304
Zaitsev M, Maclaren J, Herbst M. Motion artifacts
in MRI: a complex problem with many partial
solutions. J Magn Reson Imaging. 2015; 42(4):
887–901
89
| 09.11.19 - 01:53
5 Nuclear Medicine
Jonathon A. Nye, James R. Galt, and John N. Aarsvold
Introduction
Nuclear medicine imaging comprises planar projec-
tion and tomographic image acquisition techniques
to capture the distribution of radiolabeled
substances. Acquisitions includes single time point
imaging, gated and time-series imaging to map
radiotracer biodistributions for evaluation of
normal/abnormal function. Data collection is a
photon starved process, that is, the total number of
photons collected is substantially lower than any
other X-ray-based modality leading to images of
high noise. Moreover, attenuation and scatter within
the patient and detector coupled with an imperfect
photon detection process (detector energy resolu-
tion, collimation, and event processing) degrade the
image contrast and spatial resolution. Evaluation of
these factors, and improvements when possible, is
key to the pursuit of improved image quality.
Elements of good image quality in nuclear medicine
include absence of image distortion and a noise
distribution that is consistent with the expected
radiotracer uptake.
90
Common Image Quality
Problems
● Patient motion: degrades spatial resolution due
to object motion blurring.
● Improper attenuation correction: leads to
hypo- or hyperintense regions due to misalign-
ment between the transmission and emission
data in the reconstruction process.
● High-Z materials: oral/intravenous contrast
and implanted metallic devices are opaque
to X-rays used in positron emission tomog-
raphy/computed tomography (PET/CT)
systems to collect transmission data leading
to artifacts that propagate to the emission
image.
● Truncation: tissue lying outside the camera field
of view (FOV) that is not incorporated properly
into the reconstruction process.
| 09.11.19 - 01:53
5.1 Case 1: Degraded
Resolution of a Whole-Body
Planar 99mTc Methylene
Diphosphonate Image
5.1.1 Background
● Patient suspected of having neoplastic disease.
● Delayed phase skeletal images obtained approxi-
mately 2 hours post injection.
● Anterior/posterior planar contiguous bone
imaging protocol with low-energy high-
resolution (LEHR) collimators.
● Counts from the first images starting at the head
totaled approximately 2.3 million.
5.1 Case 1
5.1.2 Findings
● Counts are sufficient in each view (▶ Fig. 5.1).
● A large visual difference in image resolution is
observed between the anterior and posterior
planar images.
● The anterior image is of unacceptable diagnostic
quality.
5.1.3 Discussion
Review of the acquisition setup showed that the
anterior camera head was positioned further from
the patient than the posterior camera head. The
consequence was a sharp degradation in image
quality observed in the anterior camera due to the
strong dependence of resolution on the collimator
Fig. 5.1 Anterior (left) and posterior
(right) whole-body 99mTc methylene
diphosphonate images.
91
| 09.11.19 - 01:53

Nuclear Medicine

distance from the source. The loss of resolution with

collimator distance is a linear relationship with
distance from the camera as shown in ▶ Fig. 5.2.
Although the distance has a large effect on
resolution, count rate is not strongly dependent

Source-to-collimator distance
on the source distance from the collimator. The
counts collected at a single point decrease by
1/(source-to-collimator distance)2; however, the
number of collimator holes that permit passage
of photons is proportional to (source-to-
collimator distance)2.1-3 Therefore, the total
number of counts, represented by the area under
the curves in ▶ Fig. 5.2 is essentially the same
at all source distances with a parallel-hole
collimator. This relationship applies to point, line,
and uniformly distributed sources.

5.1.4 Resolution
Proper placement of the gamma camera heads
relative to the patient is the responsibility of the
camera operator. There are no software correc-
tions to sharpen a blurred image, therefore the
operator should rescan the patient with the
Fig. 5.2 Resolution of a point source (e.g., point spread
gamma camera heads placed at the proper function) versus distance from the face of a parallel-
distances from the patient. A satisfactory bone hole collimator. The total counts are the same for all
scan is shown in ▶ Fig. 5.3 with the anterior distributions.
camera closer to the patient.

92
| 09.11.19 - 01:53

5.1 Case 1

Fig. 5.3 99mTc methylene diphospho-

nate anterior and posterior images
after adjustment of camera head
position.

93
| 09.11.19 - 01:53
Nuclear Medicine
5.2 Case 2: Effect of Positron
Range on Image Quality and
Resolution
5.2.1 Background
● A rubidium-82 (82Rb) chloride PET study is
performed to assess myocardial perfusion.
● Following the resting exam, an fluorodeoxyglu-
cose ([18F]FDG) viability study is performed to
assess myocardial metabolism.
● Both studies are performed in the resting state
on the same scanner with the same reconstruc-
tion parameters.
5.2.2 Findings
● The resting [82Rb]CI and [18F]FDG viability
studies are of excellent quality exhibited by the
high contrast (> 2:1) between the myocardium
and blood pool.
● There is a marked difference in resolution
between the [18F]FDG and [82Rb]CI images,
where the [82Rb]CI image appears to be blurred
relative to the [18F]FDG study.
5.2.3 Discussion
A resting/viability protocol is a good example of
how positron range affects image quality.
Because both images were collected on the same
instrument using the same reconstruction
94
parameters, the difference in image quality can
be attributed entirely to the position range effect
(▶ Fig. 5.4). Note that some differences in image
contrast exist due to the biochemical processes
that govern the uptake and distribution of [82Rb]
CI and [18F]FDG.
▶ Fig. 5.5 details three main factors that affect
image resolution in PET imaging: detector size,
noncollinearity of the annihilation photons and
position range.4,5 Briefly, resolution is approxi-
mately proportional to half the detector size for
annihilation events originating in the center of
the FOV and approximately equal to the detector
size at the periphery. Noncollinearity in the anni-
hilation photons occurs because of a small
amount of residual momentum remaining at the
time of positron annihilation. Since the PET
system assumes collinear photons, the line along
which the system assigned the coincidence event
is in error compared to the true annihilation loca-
tion. This error in positioning increases with FOV
diameter. Lastly, the energy of positrons from
beta decay have a continuous spectrum from zero
to a maximum energy. As a result of this energy
distribution, the range of a position in tissue can
be described by an exponential function. The
higher the maximum energy, the larger the posi-
tron range and distance of the annihilation event
from the decay origin. This disparity between
location of the radiotracer and annihilation event
contributes to resolution loss independently of
the detector design. Rb-82 has maximum positron
energy of 3,400 keV compared to F-18 at 635 keV.
Therefore, all things the same, the resolution
achieved with Rb-82 will be inferior to that of F-
18-labeled compounds.
| 09.11.19 - 01:53

5.2 Case 2: Effect of Positron Range on Image Quality and Resolution

Fig. 5.4 Resting [82Rb]RbCl and [18F]FDG positron emission tomography images oriented along the short, vertical, and
horizontal axis.

95
| 09.11.19 - 01:53

Nuclear Medicine

Fig. 5.5 Illustration of resolution degrading factors is the positron emission tomography detection process.
The three main factors are noncollinearity of the annihilation photons, detector size, and position range.

5.2.4 Resolution algorithms now model the positron physics

(among other resolution degrading factors) in
The resolution difference between [82Rb]CI and the iterative reconstruction process, which can
[18F]FDG in this example is normal and attrib- compensate for some of this effect.
uted to positron range. Newer reconstruction

96
| 09.11.19 - 02:54

5.3 Case 3: Standardized Uptake Value in Positron Emission Tomography

5.3 Case 3: Standardized 5.3.2 Findings

Uptake Value in Positron ● Although the ROIs are drawn in the same
Emission Tomography location for each reconstruction, extracted

(Noise Bias) information differs.

5.3.1 Background
● Three difference whole-body [18F]FDG PET
reconstructions are presented.
● Regions of interest (ROIs) are drawn in the liver,
which is assumed to be uniform.
● Standardized uptake value (SUV) mean of the
liver is a semiquantitative indicator of scanner
calibration and may be checked periodically to
ensure correct SUV calculation.
● The three images are reconstructed with the
same iterative algorithm but with bed times of
30, 60, and 90 seconds.
● ▶ Fig. 5.6 reports the SUV mean and max
results.
5.3.3 Discussion
The SUV is calculated by normalizing the activity
concentration in the reconstructed image to the ratio
of the administered activity and body weight.6,7
h i
Bq
image mL
SUV ¼
body weight ½g  administered activity ½Bq
It has meaning similar to the pharmacological
concept of distribution volume. That is, if the
administered activity is uniformly distributed
throughout the body, the SUV would be 1.0 g/mL

Fig. 5.6 A whole-body [18F]FDG coronal slices reconstructed with imaging durations of 30, 60, and 90 seconds per
bed position. The max and mean standardized update value are reported for a region of interest placed in the liver
(red circles).

97
| 09.11.19 - 01:53
Nuclear Medicine
everywhere. This concept is independent of the
amount of FDG administered and is a method for
removing variation between patients of [18F]FDG
distribution that are related to administered
amount and body weight. Assessment of the mean
SUV in a large uniform area, such as the liver, can
be a good marker for detecting potential imaging
problems. In uniform normal liver, the mean SUV
has been shown to vary with camera manufacturer
but have an expected value ranging between 1.8
and 2.3 g/mL.
Clinically, the maximum SUV, not the mean, is
used in the reporting and staging of cancer with
[18F]FDG. SUVmax is the brightest (or highest)
voxel in the ROI and has been shown to be a
better predictor of outcomes. A number of factors
affect the accuracy of SUV including presence of
body fat, patient diet/fasting, reconstruction
parameters, scan duration, partial volume effects,
and others. In this case, we illustrate the change
in SUV on scan duration by reducing the number
of counts used in each reconstruction while
98
maintaining all other variables constant. As the
scan duration per bed is reduced, the noise in the
reconstructed images increases. Noise is a high-
frequency component in the image; therefore,
SUVmax will always increase when noise
increases. This can be thought of as a noise bias,
where the mean value is less sensitive. Although
the SUV mean is robust in the liver, it is less
reproducible than SUV max in smaller features
such as lesions.
5.3.4 Resolution
Image quality is largely subjective and based on
the preference of the interpreting radiologist or
nuclear medicine physician. It is critical that the
patient preparation and PET camera used to collect
data be kept the same for each scan in a patient's
cancer assessment. Therefore, any bias related to
the instrument hardware or reconstruction is
largely kept constant throughout a patient’s initial
staging and clinical follow-up.
| 09.11.19 - 01:53

5.4 Case 4: The Impact of Attenuation Correction in PET

5.4 Case 4: The Impact of height of the liver in the attenuation-corrected

Attenuation Correction in PET
5.4.1 Background
● An [18F]FDG PET/CT image is collected as part of
cancer staging.
● The overall image is unremarkable and consid-
ered of good technical quality.
5.4.2 Findings
● A photopenic area is observed at the interface
between the lung and liver in the attenuation-
corrected [18F]FDG PET (▶ Fig. 5.7a).
● The height of the liver in the PET scan does not
match to that in the CT scan (▶ Fig. 5.7a–c).
● The nonattenuation-corrected (NAC) scan shows
liver activity in areas that the attenuation-
corrected scan does not, suggesting these photo-
penic areas are not physiological (▶ Fig. 5.7d).
5.4.3 Discussion
The photopenic area observed in ▶ Fig. 5.7a is not
present in the NAC images in ▶ Fig. 5.4b. Also, the
images and CT are the same suggesting that the
attenuation correction was not accurate in that
interface region. This attenuation-related artifact
is common in whole-body [18F]FDG PET and is
caused by a difference in temporal resolution
between the acquisition of the PET and CT data.8
The result is that the PET represents a time-aver-
aged position of all structures that move during
the respiratory cycle while the CT captures a snap-
shot at a specific phase of the respiratory cycle.
This leads to a mismatch in position of structures
within the thoracic cavity, more easily observed
where the motion extent is the greatest, near the
lung and liver boundary.
Attenuation correction is the largest data correc-
tion step in the image reconstruction process. For a
set of annihilation photons that originate at center of
the body and travel to opposing detectors in the
coronal plane of ▶ Fig. 5.4, the correction factor can
be as large as 18 × for a total of 30 cm tissue. If the
amount of tissue traversed by these photons were
underrepresented by 40%, such as the presence of
lung instead of liver tissues, then attenuation correc-
tion factor applied would be 5.5 or 3 times
lower. An undercorrection in attenuation of this
magnitude will alter image contrast leading to
photopenic regions similar that described in
▶ Fig. 5.7.

Fig. 5.7 (a) Attenuation corrected coronal [18F]FDG positron emission tomography (FDG PET) slices showing
a photopenic area at the interface between the liver and lungs. (b) Nonattenuation-corrected coronal slice
with no photopenic regions. (c) Corresponding coronal CT used for attenuation correction of (a). (d) Attenuation-
corrected coronal FDG PET using an average CT protocol that better accounts for the liver position in the PET
data.

99
| 09.11.19 - 01:53

Nuclear Medicine

5.4.4 Resolution practice. Attenuation-related artifacts also lead to

Methods have been proposed to correct for the
motion of the diaphragm in attenuation correction
but these techniques usually add to the CT radia-
tion dose and are not widely available in clinical
scatter correction problems because this also relies
on an accurate CT. The common approach is to
“read around” the artifact but take caution if a
lesion were present in the photopenic area.

100
| 09.11.19 - 01:53
5.5 Case 5: Iterative Reconstruction and Choosing the Number of Iterations
5.5 Case 5: Iterative
Reconstruction and Choosing
the Number of Iterations
and Subsets
5.5.1 Background
● A new single-photon emission computed tomog-
raphy (SPECT) system is being installed and this
camera will perform, among other studies, 99mTc
hexamethylpropyleneamine oxime (HMPAO)
studies for assessment of cerebral blood flow.
● The imaging center has collected its first patient
and is evaluating the iterative reconstruction
parameters to determine suitable iterations
and subsets.
5.5.2 Findings
● A series of image reconstructions were made
with varying iterations and subsets (▶ Fig. 5.8).
● Attenuation correction was performed with the
Chang method and mu-value of 0.125/cm.
● A Butterworth smoothing kernel was applied
with a power 10 and cutoff of 0.7 cycles/mm.
5.5.3 Discussion
Iterative reconstruction begins with a guess
image, typically a uniform image as not to bias
the result, followed by a simulation of the scan-
ning process. The scan simulation process
includes a comparison in projection space of the
guess data with the measured projection data. A
correction image is made from this comparison
and used to update the guess image. This process
is repeated several times until the simulated
guess image best approximates the measured
projection data. Each pass through the scan
simulation process and update of the guess is
called an iteration. Note that each iteration
increases the resolution and contrast of the final
image but also the noise. The primary advantage
of iterative reconstruction is the ability to model
aspects of the scanning process, such as attenua-
tion, scatter, collimation, and other features that
affect the image appearance. A disadvantage of
iterative reconstruction is the lengthy computa-
tional time needed to run these simulations.
The iterative reconstruction process can be sped
up by dividing the projection space into subsets,
for example, simulation of 16 projections (1 of 8
subsets) out of a 128-projection acquisition.
Although only a few projections are simulated, the
scan simulation process occurs much faster and
the entire image is updated. The result is that the
guess converges to the measure projection data
faster than simulation of the entire projection
space at once. Two iterations with six subsets
result in roughly the same image contrast as six
iterations with two subsets, and this is computa-
tionally faster (▶ Fig. 5.9).
5.5.4 Resolution
Iterative reconstruction provides many advan-
tages such as modeling the physical properties of
the acquisition system that can lead to more
accurate images.9 Each iteration drives the guess
image closer to convergence but at the cost of
increasing noise. It is common to add a filter to
control for the increase in noise. The most
appropriate clinical reconstruction will depend
on the physician preferences and the exercise
shown in ▶ Fig. 5.5a is commonly done to make
those decisions.
101
| 09.11.19 - 01:53

Nuclear Medicine

Fig. 5.8 99mTc hexamethylpropyleneamine oxime images created by varying the number of iterations and subsets. Note
that the resulting contrast is similar when the iterations × subset product is the same (images along the diagonal). At
high iterations and subsets, the Butterworth filter dominates the image noise and contrast.

102
| 09.11.19 - 01:53

5.5 Case 5: Iterative Reconstruction and Choosing the Number of Iterations

Fig. 5.9 Illustration of an ordered subset iterative reconstruction process. The initial guess image is projected by the
scan simulation process. The simulated guess projections are then compared to the measured projections. A correction
image is created and used to update the guess. The process is repeated (an iteration) until the simulated scan
projections of the guess are a good approximation of the measured projections.

103
| 09.11.19 - 01:53

Nuclear Medicine

5.6 Case 6: The Effects of Image 5.6.3 Discussion

Smoothing
5.6.1 Background
● A myocardial perfusion imaging study was
performed with 99mTc sestamibi.
● The reconstructed transaxial images show a high
amount of noise than typically observed
(▶ Fig. 5.10).
5.6.2 Findings
● The counts per projection and the number of
projection samples are consistent with
published guidelines.
● Review of the reconstruction protocol shows
that iterative reconstruction was performed with
8 iterations and 10 subsets (▶ Fig. 5.11).
● Images were not corrected for attenuation.
All nuclear medicine imaging systems employ
image smoothing to improve contrast through
the reduction of statistical noise or to enhance
edges to detect boundaries. Filtering can be
performed in either the frequency domain,
commonly integrated into the reconstruction
process, or the spatial domain post reconstruc-
tion. An image can be described in terms of
summation of different frequencies, where high-
frequency components contain edge information
(e.g., air–tissue boundary) and low-frequency
components describe slow varying features
within structures. The application of a filter
either removes or modifies the frequency
components that are a part of an image, thereby
changing its appearance.
A common filter in general nuclear medicine is
the Butterworth filter, applied in frequency space,
used to remove high-frequency components but
preserve low-frequency components. Two parame-

Fig. 5.10 99mTc sestimibi myocardial perfusion images in the transaxial plane.

Fig. 5.11 A representative transaxial from a 99mTc sestamibi myocardial perfusion exam processed without filtration and
with Butterworth filters of 0.4/cm cutoff and power of 2, 5, and 20.

104
| 09.11.19 - 01:53
ters specify the filter behavior: the cutoff fre-
quency and power. The maximum cutoff frequency
is 0.5 pixels/cycle (e.g., Nyquist criterion) and
lowering the cutoff frequency will remove high-
frequency components improving low-contrast
features. The power describes how fast the filter
reaches the cutoff value and higher power
preserves more edge information.10
5.6 Case 6: The Effects of Image Smoothing
5.6.4 Resolution
The addition of a Butterworth filter to remove high-
frequency components from the image substantially
improves the image contrast. The trade-off is loss of
edge information and blurring of boundaries but, if
not overdone, this is acceptable given the improve-
ment in low-contrast resolution.
105
| 09.11.19 - 01:53
Nuclear Medicine
5.7 Case 7: Choosing the
Correct Acquisition Image
Matrix Size
5.7.1 Background
● Three-hour delayed 99mTc methylene diphospho-
nate (MDP) SPECT data were acquired to better
evaluate location and extend of disease.
● Single anterior-posterior projection and sagittal/
transaxial SPECT are evaluated.
5.7.2 Findings
● The images are of poorer diagnostic quality than
typical (▶ Fig. 5.12).
● The administered activity and counts per projec-
tion were normal.
● The acquisition settings were: low-energy high-
resolution (LEHR) collimator, 180-degree orbit,
elliptical orbit, 120 stops of 20 second each,
64 × 64 acquired projection matrix.
● The SPECT data was reconstructed with
an ordered-subset expectation maximization
(OSEM) algorithm with 6 iterations and
16 subsets that incorporates modeling of
the collimator resolution.
Fig. 5.12 Anterior-posterior projection and single-photon emission computed tomography sagittal and transaxial slices
from a 99mTc methylene diphosphonate bone study.
106
5.7.3 Discussion
A number of variables can cause degraded resolu-
tion of the reconstructed images including the
choice of collimator, camera positioning relative to
the patient, and reconstruction parameters.1 For a
typical SPECT system, the expected resolution with
a LEHR collimator is approximately 6 to 8 mm at a
distance of 10 cm from the collimator surface. Of
critical importance is to appropriately sample the
imaging space in order to utilize the capable reso-
lution of the camera. The planar projections used
to reconstruct a SPECT image consist of discrete
samples (pixels) covering the useful FOV. Choice of
acquired projection matrix will affect the resolu-
tion of the downstream processes, such as SPECT
reconstruction. The rule of thumb is to set the
sampling distance (pixel size) to one-third of the
camera resolution. For example, a camera with an
extrinsic resolution of 8 mm and square FOV of
350 × 350 mm should have 131 samples (pixels) in
each direction. SPECT instruments offer matrix
sizes scaled in powers of 2 (64, 128, 256, 1024)
and choosing 128 × 128 would provide adequate
sampling in this example. It should be noted that
angular sampling must also be set correctly and
can be estimated from relationships given in the
references.
In ▶ Fig. 5.12, the FOV is 300 mm and a matrix size
to 64 × 64 is not enough pixels to satisfy the
| 09.11.19 - 01:53

5.7 Case 7: Choosing the Correct Acquisition Image Matrix Size

sampling rule. This undersampling in the projection 5.7.4 Resolution

space leads to loss of resolution observed in
▶ Fig. 5.13 when comparing the 64 and 128 matrix
projections and SPECT reconstructions. Note that the
noise is increased in the 128 × 128 matrix as there
are fewer counts per pixel for the same total counts.
Increasing the matrix size to 256 × 256 or higher
makes no improvements on the system’s ability to
resolve structures and further increases noise.
Increasing the matrix size of the planar projections
from 64 × 64 to 128 × 128 fully utilizes the available
system resolution resulting in improved spatial
resolution and image quality. Applying a Butter-
worth filter with a high power may improve the
noise texture of data collected with the 128 × 128
matrix with limited loss is spatial resolution.

Fig. 5.13 Anterior-posterior projection and single-photon emission computed tomography slices in the sagittal and
transaxial plan from projection data acquired with a 64 × 64 matrix (top) and 128 × 128 matrix (bottom). Each acquisition
has the same total counts.

107
| 09.11.19 - 01:53
Nuclear Medicine
5.8 Case 8: Assessing Patient
Motion in Myocardial Perfusion
Imaging
5.8.1 Background
● Myocardial perfusion imaging with 99mTc
sestamibi.
● The short axis images are of poor in quality
(▶ Fig. 5.14).
5.8.2 Findings
● There is some distortion in the short axis heart
size, particularly along the septal and inferior
wall.
● Inspection of the planar images shows axial
motion in the table direction in some but not
all projections (▶ Fig. 5.15).
5.8.3 Discussion
The lengthy imaging time of myocardial perfu-
sion SPECT makes this data acquisition process
particularly susceptible to patient motion.11
Patient motion such as contractile cardiac, respi-
ratory action or movement on the table can
degrade image resolution and introduce artifacts
that often appear as structural distortions.
Cardiac and respiratory action can be addressed
using gating techniques but patient body motion
on the table cannot be addressed using these
tools. Patient motion can be minimized through
Fig. 5.14 Reoriented short axis slices of a 99mTc sestamibi myocardial perfusion study. Note the distortion in the superior
and septal walls.
108
good technologist instructions or using light
restraints. The step-wise collection of projection
data provides another means to assess motion by
playing them back in a loop to view changes in
the heart’s position. Up and down shifts indicate
motion along the superior/inferior direction (e.g.,
table direction), whereas lateral/medial shifts are
more difficult to assess due the angular sampling
of the SPECT acquisition.
Tools are available from several manufacturers
to minimize patient motion artifacts. These tools
permit manual (or automatic) shifts of the
projection data. For example, in ▶ Fig. 5.15,
shifting projections 17 and 23 approximately two
pixels would correctly align the heart with the
neighboring projections. ▶ Fig. 5.16 shows the
same data in ▶ Fig. 5.14 after the projection data
was motion corrected for the superior/inferior
movement. Note the improved image quality and
remediation of structural distortion compared to
the motion-corrupted reconstruction. A note of
caution is that shifting projections can do more
harm than good and automated routines should
be reviewed for accuracy before sent to recon-
struction.
5.8.4 Resolution
Myocardial perfusion imaging performed with
SPECT should be examined for possible motion
corruption. This can be easily assessed by viewing
playback of the planar projection data for changes
in the heart position. Tools that correct for volun-
tary motion work well but should be reviewed for
accuracy if used in an automated state.
| 09.11.19 - 02:56

5.8 Case 8: Assessing Patient Motion in Myocardial Perfusion Imaging

Fig. 5.15 A sample of planar projections used in the reconstruction of ▶ Fig. 5.8 (original panel). The red arrows show a
dip in the heart position indicating the patient moved down in the direction of the table. Following motion correction
(motion corrected panel), the heart is aligned in all projections.

Fig. 5.16 Reoriented reconstructed short axis slices of a 99mTc sestamibi myocardial perfusion study following motion
correction of the planar projections.

109
| 09.11.19 - 01:53
Nuclear Medicine
5.9 Case 9: Bremsstrahlung
Imaging of 90Y Microspheres
Liver Embolization
5.9.1 Background
● A patient workup for intra-arterial resin micro-
sphere radioembolization of a hepatic lesion.
● They have received a 99mTc microaggregated
albumin (MAA) pretreatment planning scan.
● Post radioembolization SPECT/CT data collection
of 90Y bremsstrahlung photons was performed.
● Data are fused with contrast enhance MRI to
confirm the distribution of trapped micro-
spheres is in the tumor vasculature.
5.9.2 Findings
● Setup of the 90Y SPECT camera includes use of a
medium-energy collimator.
● There is no photopeak to window, as with 99mTc
imaging. A 35-keV energy window centered on
108 keV was selected.
● Data were collected with 128 × 128 matrix and
reconstructed with OSEM and no attenuation
correction.
5.9.3 Discussion
90Yemits a high-energy beta particle (2.28 MeV
maximum) that undergoes bremsstrahlung
Fig. 5.17 Photon energy spectrums from unshielded 99mTc (a) and 90Y (b). The photopeak energy at 72–88 keV is a Pb
characteristic X-ray from photoelectric interactions with the lead collimator.
110
interactions in soft tissue as it slows down
resulting in a wide spectrum of photons with a
maximum energy of approximately 2.3 MeV. Since
there is no distinct photopeak, as typical for
gamma camera imaging, placement of an energy
window is not trivial (▶ Fig. 5.17). The placement
and width of the imaging window have been
suggested to be around 108 keV with a 35 keV
width when using a medium-energy collimator.
Conventional energy discrimination is not possible
with a continuous photon spectrum; therefore, the
contribution of scatter is difficult to quantify in the
image. In addition, since the incident photons
range in energy well above 108 keV, there is
considerable septal penetration and scatter within
the collimator as well as the camera housing that
further contribute to degraded image quality. With
all of these effects, it is still possible to obtain good
quality images of 90Y biodistribution post emboli-
zation with SPECT as shown in ▶ Fig. 5.18. Further
advancements in SPECT quantification have
worked toward verifying the fraction of adminis-
tered activity in the liver but the physical
complexity increases as corrections for both
attenuation and scatter in soft tissue depend on
photon energy.
5.9.4 Resolution
Imaging of bremsstrahlung photons can be
performed with good image quality but careful
consideration is needed when selecting an energy
window and collimator as these factors affect
image quality.
| 09.11.19 - 01:53

5.9 Case 9: Bremsstrahlung Imaging of 90Y Microspheres Liver Emboliza

Fig. 5.18 99mTc microaggregated albumin (MAA) planning (middle panel) and 90Y bremsstrahlung (bottom panel) single-
photon emission computed tomography images fused with MR. Patient is a 71-year-old male. 4 mCi 99mTc MAA and
28.6 mCi 90 SIR-Sphere administrations.

111
| 09.11.19 - 01:53
Nuclear Medicine
5.10 Case 10: Degraded Image
Quality from an Improper
Collimator
5.10.1 Background
● Patient screened for bone metastases with
known cancer.
● 25 mCi of 99mTc MDP was administered.
● Whole-body planar imaging was performed
about 45 minute after MDP administration
(▶ Fig. 5.19).
5.10.2 Findings
● Images are of inferior quality compared to
previous patient images (▶ Fig. 5.20).
● The image resolution appears worse than typical,
which is more evident in the posterior planar
view.
5.10.3 Discussion
Review of the protocol and patient setup
revealed a technical error where the medium-
energy collimators were used during imaging.
The technologist operating the camera had
previously performed imaging with 111In (photo-
112
peaks: 171 keV, 245 keV) and medium-energy
collimators. LEHR collimators are most
commonly used with 99mTc-labeled radiophar-
maceuticals and are specifically designed to opti-
mize resolution and sensitivity of 140 keV
photons from 99mTc. Compared to Low energy
high resolution (LEHR), medium-energy collima-
tors typically have larger holes and thicker septa
in order to maintain sensitivity and prevent
resolution degradation from septal penetration
when imaging with photon energies above
approximately 200 keV.
The consequence of using medium-energy
collimators with 99mTc is a loss of resolution due to
the larger collimator hole aperture as illustrated in
▶ Fig. 5.21. The wider angle of acceptance allows
more photons to reach the detector crystal from
oblique angles. This leads to a degraded contrast
and spatial resolution for objects located at the
same distance compared to imaging with a LEHR
collimator.
5.10.4 Resolution
Loss of resolution from this technical error
cannot be resolved with post-processing. The
technologist should switch from medium-energy
collimators to low-energy collimators and
reimage the patient. A satisfactory bone scan
with the proper collimators is shown in
▶ Fig. 5.10.
Fig. 5.19 Repeat bone scan collected
for evaluation of bone metastases
(right). Prior bone scan on the same
camera from a different patient (left).
There is a degradation in image reso-
lution from left to right that is also
visible in the anterior view (not shown).
Fig. 5.20 Medium-energy collimators
(left) and a repeat scan with low-
energy high-resolution collimators
(right).
| 09.11.19 - 01:53

5.10 Case 10: Degraded Image Quality from an Improper Collimator

Fig. 5.21 Illustration of the effect of

resolution for a medium- and low-
energy collimator for a source at the
same distance from the collimator. The
measure of point spread function of a
point source at the same distance is
wider for the medium-energy colli-
mator compared to the low-energy
collimator. Widening of the point
spread function results in degraded
resolution and contrast.

113
| 09.11.19 - 01:53

Nuclear Medicine

5.11 Review Questions c) Detector size

5.11.1 Case 1: Degraded
Resolution of a Whole-Body Planar
99mTc Methylene Diphosphonate
Image
1. A quality control source of Technetium-99 m
(99mTc) yielding 1200 cps is placed 10 cm from
the face of a gamma camera with a parallel-
hole collimator. If the source is move to 20 cm
from the face of the gamma camera, what is
the expected count rate?
a) 1800 cps
b) 1200 cps
c) 800 cps
d) 600 cps
2. The parallel-hole collimator having 1 cm length
holes is replaced with another parallel-hole
collimator of 1.5 cm length holes of the same
septal thickness. This change will result in a
reduction of what performance characteristic?
a) Sensitivity
b) Resolution
c) Magnification
d) Energy resolution
d) Depth of interaction
5.11.3 Case 3: Standardized Uptake
Value in Positron Emission
Tomography (Resolution and Noise)
5. Calculate the standardized update value
for patient weighing 80 kg who is administered
10 mCi of [18F]FDG with a decay corrected
measured lesion concentration of 300 nCi/mL.
a) 2.4
b) 3.2
c) 4.1
d) 5.3
6. The mean liver SUV from an [18F]FDG oncology
PET/CT study was measured to be 1.6. It was
later discovered that the patient weight was
incorrectly documented during the exam as 180
lb and should have been recorded as 230 lb.
What is the mean liver SUV with this revised
weight?
a) 1.6
b) 1.8
c) 2.0
d) 2.2

5.11.2 Case 2: Effect of Positron

Range on Image Quality and
Resolution
3. The system response function, including all
resolution-degrading factors of detector size,
noncollinearity, and positron range, is meas-
ured by imaging what type of object?
a) Point source
b) Rotating rod source
c) Large uniform phantom
d) An anthropomorphic phantom
4. Which resolution-degrading aspect of the PET
image does not depend on the instrument’s
construction?
a) Noncollinearity
b) Positron range
5.11.4 Case 4: The Impact of
Attenuation Correction in PET
7. Including attenuation correction, which other
data correction process in PET relies on accu-
rate measurement of the attenuation map?
a) Normalization
b) Scatter
c) Randoms
d) Well-counter calibration
8. What is the primary interaction process of
511 keV annihilation photons in the body?
a) Coherent scattering
b) Compton scattering
c) Photoelectric absorption
d) Pair production

114
| 09.11.19 - 01:53

5.11 Review Questions

5.11.5 Case 5: Iterative 5.11.7 Case 7: Choosing the

Reconstruction and Choosing the Correct Acquisition Image
Number of Iterations and Subsets Matrix Size
9. Which reconstruction approach is not based 13. What is the pixel size of a planar gamma
on an iterative algorithm? acquisition with a 25-cm FOV collected with a
a) Maximum likelihood expectation 128 × 128 pixel matrix?
maximization a) 1.20 mm
b) Conjugate gradient minimization b) 1.95 mm
c) Algebraic reconstruction technique c) 2.32 mm
d) Filtered backprojection d) 3.21 mm

10. What is an advantage of using filtered backpro- 14. When switching from a 64 × 64 to a 128 × 128
jection over iterative reconstruction routines? matrix, how many more counts are need
a) Increased speed to maintain the same noise properties
b) Reduced noise (× = times)?
c) Improved spatial resolution a) 1 ×
d) Resolution modeling b) 2 ×
c) 3 ×
5.11.6 Case 6: The Effects of d) 4 ×

Image Smoothing
5.11.8 Case 8: Assessing Patient
11. Which image filtering type is commonly
applied in the image space domain after
Motion in Myocardial Perfusion
reconstruction?
Imaging
a) Gaussian 15. When projection data are organized into a
b) Butterworth sinogram, patient motion along the table direc-
c) Shepp–Logan tion leads to as what appearance in the sinus-
d) Ramp oidal information?
a) Area of high counts
12. Increasing the Butterworth cutoff frequency b) Data discontinuities
toward the Nyquist limit changes which c) Complete loss of data
characteristic of image quality? d) No change can be observed
a) Reduces noise
b) Improves low contrast 16. How does patient motion change image
c) Increases resolution resolution?
d) Introduces aliasing a) Degrades
b) No change
c) Improves

115
| 09.11.19 - 01:53

Nuclear Medicine

5.11.9 Case 9: Bremsstrahlung References

Imaging of 90Y Microspheres Liver [1] Naddaf SY, Collier BD, Elgazzar AH, Khalil MM. Technical er-

Embolization rors in planar bone scanning. J Nucl Med Technol. 2004; 32

17. What photon interaction process produces the
Pb characteristic X-rays commonly observed in
a gamma camera energy spectrum?
a) Photoelectric
b) Compton
c) Coherent
d) Pair production
18. What is a common energy window width used
in 99mTc radiopharmaceutical imaging?
a) 10–15%
b) 15–20%
c) 20–25%
d) 25–30%
5.11.10 Case 10 Degraded Image
Quality from an Improper
Collimator
19. What collimator is appropriate for performing
a planar gallium-67 (Ga-67) citrate study?
(Ga-67 has a photopeak at 93 keV (37%), 187
keV (20.4%), and 300 keV (16.6%)).
a) Low-energy high-resolution
b) Low-energy general-purpose
c) Medium-energy
d) High-energy
(3):148–153
[2] Cherry SR, Sorenson JA, Phelps ME. The gamma camera: basic
principles. In: Cherry SR, Sorenson JA, Phelps ME, eds. Physics
in Nuclear Medicine 4th ed. Philadelphia: W.B. Saunders;
2012:195–208
[3] Cherry SR, Sorenson JA, Phelps ME. The gamma camera: per-
formance characteristics. In: Cherry SR, Sorenson JA, Phelps
ME, eds. Physics in Nuclear Medicine. 4th ed. Philadelphia,
PA: W.B. Saunders; 2012:209–31
[4] Cherry SR, Sorenson JA, Phelps ME. Positron emission tomog-
raphy. In: Cherry SR, Sorenson JA, Phelps ME, eds. Physics in
Nuclear Medicine. 4th ed. Philadelphia, PA: W.B. Saunders;
2012:307–43
[5] Votaw JR. The AAPM/RSNA physics tutorial for residents.
Physics of PET. Radiographics. 1995; 15(5):1179–1190
[6] Kinahan PE, Fletcher JW. Positron emission tomography-
computed tomography standardized uptake values in clinical
practice and assessing response to therapy. Semin Ultrasound
CT MR. 2010; 31(6):496–505
[7] Thie JA. Understanding the standardized uptake value, its
methods, and implications for usage. J Nucl Med. 2004; 45
(9):1431–1434
[8] Blodgett TM, Mehta AS, Mehta AS, Laymon CM, Carney J,
Townsend DW. PET/CT artifacts. Clin Imaging. 2011; 35(1):
49–63
[9] Tong S, Alessio AM, Kinahan PE. Image reconstruction for
PET/CT scanners: past achievements and future challenges.
Imaging Med. 2010; 2(5):529–545
[10] Galt JR, Hise HL, Garcia EV, Nowak DJ. Filtering in frequency
space. J Nucl Med Technol. 1986; 14(3):152–160
[11] Burrell S, MacDonald A. Artifacts and pitfalls in myocardial
perfusion imaging. J Nucl Med Technol. 2006; 34(4):193–
211, quiz 212–214

20. Increasing the collimator hole length has what

effect on image quality?
a) Reduction in noise
b) Improved resolution
c) Increased sensitivity
d) Decreased scatter rejection

116
| 09.11.19 - 01:53
6 Ultrasound Imaging
Zheng Feng Lu
Introduction
An ultrasound image is formed using sound. Ultra-
sound has been widely used because it is portable,
low cost, and does not involve ionizing radiation;
thus, it is safe even for scanning a fetus. B-mode
ultrasound image formation is based upon three
basic assumptions:1
1. The sound travels in a straight and narrow
line called an acoustic beam. A transducer is
used both as a pulse emitter and as an echo
receiver. The transducer emits a short pulse
and then receives echo signals which are
generated by the emitted pulse propagating
and interacting with tissue within the
acoustic beam.
2. The speed of the sound propagation is assumed
to be constant at 1,540 m/s, an average speed
for soft tissues. Therefore, on the image, the
sources for echo signals are localized by the
so-called range equation:
ct
D¼
2 ●
Where t is the time delay between the pulse
emission and the echo reception, c represents
the speed of sound (i.e., 1,540 m/s is assumed
for soft tissue), and D is the depth of the echo
where it is generated. The time delay t includes
both the time it takes for the pulse to travel
down to the reflector and the time it takes for
the echo to return. Therefore, it accounts for the
factor of 2.
3. The received echo signals are amplified to
compensate for the attenuation. The strength
of the echo signals is represented by varying
brightness levels on a B-mode image.
Ultrasound images are formed using the ideal-
ized model described above. Ultrasound artifacts
occur when any or all of the assumptions are
violated.
Common Image Quality
Problems
● Speed of sound propagation: it affects spatial
resolution and the accuracy of the distance
measurements.
● Frequency: it affects spatial resolution and the
maximum depth of penetration.
● Array transducer dropouts: the deficiency most
commonly found during routine quality control
(QC) testing.
● Presets in image protocols: optimize the image
acquisition controls and maintain the consis-
tency by managing the “presets” in image
protocols.
● Acoustic window: it affects the coupling of the
transducer and patient body. A poor acoustic
window prevents the sound from being trans-
mitted to the region of interest.
● Reverberation artifacts such as comet-tail and
ring-down can provide diagnostic information.
Range ambiguity occurs when the pulse repeti-
tion frequency (PRF) is too high with the result
that echoes from the prior beam line are mispo-
sitioned in the current beam line.
● Enhancement and shadowing can provide
diagnostic information.
● Harmonic imaging is advantageous over conven-
tional B-mode imaging in generating superior
image quality, especially in the case of a “techni-
cally difficult” patient with a thick body wall.
● The performance evaluation of the ultrasound
scanner display and the reading room worksta-
tion displays is crucial in maintaining consis-
tency of image perception.
● Doppler aliasing occurs when the PRF is too low,
a control that is linked to the PRF. To reduce
aliasing, one should increase the velocity scale.
117
| 09.11.19 - 01:53
Ultrasound Imaging
6.1 Case 1: Pulse-Echo Imaging
Principle and Speed of Sound
Propagation
6.1.1 Background
Size measurement is a routine clinical application
in ultrasound imaging. The range equation is
implicitly built into pulse-echo imaging instru-
ments to localize the received echo signals. If the
sound propagation speed in the body tissue is
indeed 1540 m/s, the measured size will be accu-
rate. However, if the sound propagation speed is
less than 1540 m/s, then the longer delay in echo
return time is interpreted by longer distance; in
this case the distance in the axial direction is over-
estimated. On the other hand, if the speed is
greater than 1540 m/s, then the distance in the
axial direction is underestimated.
The misestimated sound propagation speed not
only affects the distance measurement, but also
contributes to the reduction of the spatial resolu-
tion; thus, it worsens the image quality.
6.1.2 Findings
In the following clinical case, the discontinuity seen
in the diaphragm pointed by the arrow in ▶ Fig. 6.1a
is caused by the different sound speed in the lesion.
The sound speed reduces in a fatty lesion. Therefore,
the portion of the diaphragm (pointed by the arrow)
below the lesion is shown in a deeper position in the
image. In another clinical case shown in ▶ Fig. 6.1b,
an irregular liver/diaphragm interface is seen in the
image caused by heterogeneous liver parenchyma
with fatty infiltration.
Fig. 6.1 Speed artifacts shown as the discontinuity at the liver/diaphragm interface (pointed by the arrow) caused by a
lesion (a). An irregular liver/diaphragm interface shown in the image caused by heterogeneous liver parenchyma with
fatty infiltration (b).
118
The misestimated speed of sound propagation
affects the image quality as demonstrated by the
group of pins in ▶ Fig. 6.2.
6.1.3 Discussion
The speed of sound propagation is dependent
upon the tissue properties such as density and
compressibility. For any particular material, the
higher its density and the harder it is to
compress, the faster its sound speed will be. For
example, the sound speed is faster in bone than
that in soft tissue because bone is more dense
and harder to compress than soft tissue.
▶ Table 6.1 shows the typical speed of sound
propagation in a variety of materials. In diag-
nostic ultrasound, a sound speed of 1,540 m/s is
assumed. This value represents the average speed
of sound propagation in soft tissues. Any devia-
tion from the assumed speed causes errors in size
measurements, known as speed artifacts. As the
acoustic beam focusing relies on the assumption
of the sound speed, image quality is affected by
any deviation in the actual sound speed and the
assumed sound speed by the ultrasound system
in setting up the beam former. As demonstrated
in ▶ Fig. 6.2, the beam focuses the best when the
assumed sound speed matches the actual sound
speed at 1460 m/s.
6.1.4 Resolution
It is important to recognize potential speed
artifacts by examining the acoustic pathway
and identifying regions that are suspected
to have different values for sound speed. Typically,
| 09.11.19 - 01:53

6.1 Case 1: Pulse-Echo Imaging Principle and Speed of Sound Propagation

Fig. 6.2 (a–d) A set of images of a group of pins in a urethane phantom under different settings of speed of sound
propagation. The actual speed of sound in urethane is 1,450 m/s as per manufacturer’s specification. The spatial
resolution is at its best when the speed is set at 1,460 m/s that is the closest to the actual speed in urethane material.

Table 6.1 Sound propagation speed in selected materials with 1,540 m/s as the average speed for soft tissue2
Media Lung Fat Water Liver Blood Kidney Muscle Skull bone
Speed 600 1,460 1,498 at 1,555 1,560 1,565 1,600 4,080
(m/s) 25 °C

nothing is done to correct for speed artifacts.

However, some modern ultrasound systems allow-
manual adjustment to correct for deviations in the
speed of sound in order to improve image quality. As
demonstrated in ▶ Fig. 6.2, the actual speed of sound
of the urethane-based phantom is approximately 6%
lower than 1,540 m/s.3 When the machine-assumed
speed of sound does not match the actual speed of
sound in the urethane-based phantom, the lateral
resolution deteriorates. This feature has been used
Fig. 6.3 A breast image with the speed of sound
for breast imaging. As shown in ▶ Fig. 6.3, the image
propagation set at 1,500 m/s to match the actual
quality is improved by an adjustment from the
speed of sound in breast tissue to improve image
typical sound speed, i.e., 1,540 m/s to the actual quality.
sound speed in breast tissue at 1,500 m/s.

119
| 09.11.19 - 01:53

Ultrasound Imaging

6.2 Case 2: Array Transducers Linear Array

and Sound Frequency ● A subgroup of elements is fired to form one

6.2.1 Background acoustic beam perpendicular to the face of the

Sound frequency is the number of oscillations per
unit time, determined by the source—ultrasound
transducer. For human ears, the audible sound fre-
quency ranges from 20 Hz to 20 kHz. Sound with
frequencies above 20 kHz is called ultrasound. The
higher the frequency, the better the spatial resolu-
tion in ultrasound images. On the other hand, the
lower the frequency, the deeper the ultrasound
penetrates because the ultrasound attenuation is
proportional to frequency. In medical ultrasound
imaging, the frequency typically ranges from 2 to 15
MHz with low frequency used for imaging deep
targets but with compromised spatial resolution,
and high frequency used for imaging superficial
targets with better spatial resolution. High frequen-
cies, around 50 MHz or even higher, are used for
certain specialized imaging applications, such as in
ophthalmology or in small animal imaging, with
superb spatial resolution.
An ultrasound system typically has multiple
transducers of various frequencies, each trans-
ducer having a different contact surface size (foot-
print) designed for specific clinical applications.
Modern ultrasound systems often allow the oper-
ator to choose the frequency within a range on the
same transducer, which allows for adjustment
between penetration and spatial resolution.
6.2.2 Findings
Different types of array transducers and their corre-
sponding image examples are shown in ▶ Fig. 6.4.
array.
● Subgroups of elements are fired sequentially to
form parallel acoustic beams to scan across a
region of interest (ROI).
● Rectangular shape of the image format.
● Clinical applications include small parts, vascular,
and obstetric exams (▶ Fig. 6.4a).
Curvilinear or Curved Array
● Same as linear arrays except that the elements
are aligned in a curve instead of a straight line.
● Each acoustic beam is perpendicular to the face of
the array. Since the array of elements is curved, the
acoustic beams diverge with depth which allows
for a broader coverage as the depth increases.
● Clinical applications include general abdominal,
obstetric, and transabdominal pelvic exams. In
addition, transvaginal and transrectal probes are
curved arrays (▶ Fig. 6.4b).
Phased Array
● All the elements are fired to form one acoustic
beam.
● The acoustic beam is electronically steered
across the ROI.
● The footprint is typically very small to allow for
a small acoustic window (e.g., between the ribs)

Fig. 6.4 Various types of array transducers and their corresponding image examples. (a) An image example of a linear
array transducer. (b) An image example of a curvilinear array transducer. (c) An image example of a phased array
transducer.

120
| 09.11.19 - 01:53
6.2 Case 2: Array Transducers and Sound Frequency
but the sector image format allows for a broad
field of view.
● Clinical applications include intercostal scanning
for heart, liver, or spleen (▶ Fig. 6.4c).
6.2.3 Discussion
On modern ultrasound systems, frequency can be
selected within a certain range on the same trans-
ducer. While making the frequency selection, one
must consider the fundamental trade-off between
spatial resolution and penetration, and strike an
optimal balance between these two factors
(▶ Fig. 6.5).
6.2.4 Resolution
It is possible to improve penetration at high fre-
quency with advanced technology such as coded
excitation technology. This technology plants a
code in the transmitted pulses. In return, the
received echo signals are carrying the same code;
thus, raising the ability to differentiate between
echo signals and noises. Through appropriate
coding on transmitted pulses and decoding on
received echo signals, one can improve the signal-
to-noise ratio and can, therefore, mitigate the
trade-off between better spatial resolution, which
is associated with higher frequency, and less
penetration, also associated with higher frequency.
Fig. 6.5 A phantom is scanned by
the same transducer with different
frequency selections.
121
| 09.11.19 - 01:53

Ultrasound Imaging

6.3 Case 3: Nonuniformity the wires inside the cable and cause element
dropouts. Handling transducers with care is crit-
(Array Transducer Element ical to ensuring the longevity of these array
Dropouts) transducers.

6.3.1 Background
An ultrasound array transducer contains an array
(or arrays) of composite ceramic piezoelectric
elements connected by wires enclosed in a cable
that runs to the ultrasound system via a
connector. An ultrasound transducer is vulner-
able to damage because it is handled frequently
during ultrasound imaging operation and may
easily be dropped or bumped against hard
surfaces, or its cable may be rolled underneath
the scanner wheels during transportation for
portable studies. An ultrasound transducer is
prone to image nonuniformity problems due to
any of the following defects: (1) failed crystal
elements; (2) delamination of the lens/coupling
layers on the transducer face; (3) broken wires
in the transducer cable; and (4) disruptions in
the connector. A picture of transducers with
tangled cables connected to an ultrasound
system is shown in ▶ Fig. 6.6a. Transducers are
often hung on walls, as shown in ▶ Fig. 6.6b, c.
Too much stress on a transducer cable may tear
6.3.2 Findings
Sometimes the nonuniformities are minor,
appearing as streaks along the axial direction
of the transducer, whereas some are more
prominent. ▶ Fig. 6.7 shows the images generated
by the same transducer. Nonuniformities can be
seen on the clinical image and the phantom image,
as well as on the in-air scan image.
6.3.3 Discussion
Image nonuniformity is considered the most
commonly found deficiency during routine QC
testing.4 As shown in the example in ▶ Fig. 6.8, the
transducer cable was accidentally rolled under-
neath the scanner wheel and five wires in the
cable were broken causing minor dark streaks near
the face of the transducer. The broken wires were
detected by an electronic transducer testing
device.5 The nonuniformity was much less perceiv-
able on clinical images.

Fig. 6.6 Ultrasound transducers are handled frequently during ultrasound imaging operation. Education on careful
handling of the transducers can reduce transducer failure rate. (a) Transducers should be hung properly on the system
to avoid tangles or being trapped under the wheels of the system. (b) While it is neat to hang up transducers, too much
stress on transducer cable may tear the wires inside the cable and cause element dropouts. (c) Various transducer
hanging boxes such as this one are designed to handle transducers with care by minimizing the stress on transducer
probe, its connector and cable.

122
| 09.11.19 - 02:58

6.3 Case 3: Nonuniformity (Array Transducer Element Dropouts)

Fig. 6.7 (a–c) Nonuniformities are seen in the images of this transducer (pointed by arrows) on a patient, a phantom,
and just in-air.

Fig. 6.8 Both images were obtained by the same transducer. (a) The streaks observed near the face of the transducer in
the phantom image are hard to see in the (b) clinical image.

6.3.4 Resolution before clinical applications are impacted. The image

It is important to perform periodic QC testing using
an ultrasound QC phantom to detect the presence
of any nonuniformities as early as possible in order
to monitor the condition and address the problem
uniformity test is a required QC test by American
College of Radiology (ACR) Ultrasound Accredita-
tion Program for all transducers used for clinical
applications.6

123
| 09.11.19 - 02:58

Ultrasound Imaging

6.4 Case 4: Pulse-Echo Imaging 6.4.3 Discussion

Acquisition Controls
6.4.1 Background
The pulse-echo imaging controls on an ultrasound
system affect both image data acquisition and
image display. The proper setting of these controls
is crucial for optimizing the image performance to
fulfill the clinical task.
6.4.2 Findings
We will discuss the following four main controls:
● Transmit power (▶ Fig. 6.9)
● Overall gain (▶ Fig. 6.10 and ▶ Fig. 6.11)
● Time gain compensation (TGC) (▶ Fig. 6.12)
● Dynamic range (▶ Fig. 6.13)
Four ultrasound pulse-echo imaging acquisition
parameters are explained below.
● Transmit power: This determines the amplitude
of the pulse emitted from the transducer. Higher
transmit power will yield stronger returned
echoes, thus deeper penetration depth as demon-
strated in ▶ Fig. 6.9. This control affects the acoustic
dosimetry that is typically indicated by the
mechanical index (MI) and the thermal index (TI).
● Gain (overall gain): This amplifies the echo signals
of all depths. It is often used to adjust the overall
image brightness. Unlike transmit power control,
this control does not affect the penetration depth,
as demonstrated in ▶ Fig. 6.10 and ▶ Fig. 6.11.
Neither does it affect the acoustic dosimetry.

Fig. 6.9 The ultrasound phantom images were acquired under identical instrumentation settings except the transmit
power setting at maximum in (a), − 6 dB in (b) and − 12 dB in (c). As the transmit power reduces, the image becomes
dimmer and with less penetration.

Fig. 6.10 The ultrasound phantom images were acquired under identical instrumentation settings except the overall
gain setting at 100% in (a), 70% in (b), and 60% in (c). As the overall gain reduces, the image becomes dimmer but the
penetration remains the same.

124
| 09.11.19 - 01:53

6.4 Case 4: Pulse-Echo Imaging Acquisition Controls

Fig. 6.11 The ultrasound phantom images were acquired under identical instrumentation settings except that the
overall gain setting is increased by 5 dB at each increment from 30 dB at the upper left panel to 70 dB at the lower
right panel.

Fig. 6.12 The two ultrasound

phantom images were acquired under
identical instrumentation settings
except that the time gain compensa-
tion (TGC) setting is set at the center
position in the upper panel and one of
the TGC knobs was set at its minimum
position in the lower panel.

125
| 09.11.19 - 01:53
Ultrasound Imaging
Fig. 6.13 (a) The ultrasound images were acquired under identical instrumentation settings except the dynamic range
setting at 36 dB, (b) 60 dB, and (c) 96 dB. As the dynamic range increases, the image becomes smoother but with less
contrast.
● TGC: Due to ultrasound attenuation, echoes
returning to the transducer get weaker from the
distance traveled. The longer distance it travels,
the weaker the echo signal becomes. Amplifica-
tion in the receiver, called TGC, is needed to offset
the loss due to attenuation as the depth increases.
By increasing amplification along with the depth,
TGC offers a more uniform display of the bright-
ness level throughout the field of view (FOV). TGC
can be altered by a group of sliding knobs, each of
which amplifies the echo signals from a specific
depth range. The TGC knobs are calibrated in such
a way that when all the knobs are aligned in the
center position, the image of an organ, for
example, a liver, appears with uniform brightness
at all depths as demonstrated in ▶ Fig. 6.12.
● Dynamic range: In ultrasound imaging, echo
signals are logarithmically compressed and
transformed by decibel notation, defined as
10 times the log10 of the ratio of echo signal
intensity compared to reference intensity. The
126
dynamic range indicates the range from the
smallest to the largest echo signals to which the
system can properly respond. It is described in
decibels. A smaller dynamic range setting
means a steeper gradient that provides more
contrast on the image display, increasing the
conspicuity of a low-contrast lesion. However,
the image will appear coarse. Conversely, a
larger dynamic range setting has a lower
contrast gradient, but the image will appear
smoother. This is demonstrated in ▶ Fig. 6.13.
6.4.4 Resolution
The ultrasound image acquisition controls,
discussed here, are user adjustable. Typically,
optimization of these controls is conducted
through “presets” on the system protocol and
may vary per transducer and per body part, even
for an ultrasound system of the same vendor and
same model. Therefore, understanding and
managing these controls helps maintain the
consistency of the imaging performance.
| 09.11.19 - 01:53
6.5 Case 5: Reflection (Boundary Conditions)—Reverberation Artifacts
6.5 Case 5: Reflection
(Boundary Conditions)—
Reverberation Artifacts
6.5.1 Background
A sound beam is reflected when it is incident on an
interface formed by two tissues of different acoustic
impedance. The magnitude of the reflection
depends on the acoustic impedance difference at
the interface. The smaller the acoustic impedance
difference, the less the reflected beam energy, and
the better the transmission of the sound beam will
be. Acoustic impedance is a property of the tissue
that equals to the product of tissue density and the
sound propagation speed. Therefore, the acoustic
impedance of soft tissue is very different from the
acoustic impedance of bone or gas. A good acoustic
window is a body location that allows minimal
sound reflection and optimal sound transmission,
for example, region with no bone or gas to compro-
mise sound transmission.
6.5.2 Findings
Ultrasound imaging assumes that an echo returns
to the transducer after a single reflection. Rever-
beration artifacts occur when this assumption
fails, as shown in ▶ Fig. 6.14.
6.5.3 Discussion
When two highly reflective interfaces run
parallel in the beam propagation path, it may
Fig. 6.14 Reverberation artifacts occur due to parallel
reflective interfaces in the body wall.
result in multiple reflections back and forth
between the two interfaces. The first received
echo will appear at the proper depth on the
image, but the subsequent echo signals received
from the multiple reflections between the two
interfaces will appear at deeper depths; the extra
time delays caused by multiple reflections are
interpreted by the system as representing longer
distances traveled by the sound. These additional
echo signals appear equally spaced on the image
but with decreasing intensity, as each reflection
is weaker than the prior.
Comet-tail artifact as shown in ▶ Fig. 6.15 is a
special subtype of reverberation artifact which
occurs when the two highly reflective interfaces
are too close to each other to be resolvable on the
image. As implied by its name, comet-tail artifact
looks like a comet tail with a tapering echogenic
triangle-shaped tail. It often appears in response to
a highly reflective object such as cholesterol crys-
tals in adenomyomatosis of the gallbladder or a
bullet fragment.
Ring-down artifact as shown in ▶ Fig. 6.16 is
often included as another special subtype of rever-
beration artifacts. Although it may resemble a
comet-tail artifact, the ring-down artifact is
different. Caused by resonant vibrations within the
center of a cluster of air bubbles, a ring-down arti-
fact appears as a long solid streak or series of steaks
along the sound propagation axis. In a 1985 publica-
tion,7 Avruch et al demonstrated that ring-down
never occurred with only one layer of air bubbles.
Rather, it only resonated from a fluid center trapped
within a cluster or tetrahedron of air bubbles: three
air bubbles on top and one air bubble nestled
beneath.
Fig. 6.15 Breast imaging with surgical clips showing
comet-tail artifacts.
127
| 09.11.19 - 01:53

Ultrasound Imaging

6.5.4 Resolution
The reverberation artifacts are easy to identify, as
shown in ▶ Fig. 6.14, ▶ Fig. 6.15, and ▶ Fig. 6.16.
Typically, nothing is done about it. Sometimes, a
different acoustic window is chosen to avoid severe
reverberation artifacts. Comet-tail artifacts are clini-
cally useful because small objects, such as surgical
clips, can be identified through the occurrence of
comet-tail artifacts, as shown in ▶ Fig. 6.15. Ring-
down artifact can be useful in providing diagnostic
information, for example, in case of emphysematous
Fig. 6.16 Ring-down artifact caused by air bubbles in
(gas-forming) infections and abscesses that often
the bowel.
produce ring-down artifacts.

128
| 09.11.19 - 01:53
6.6 Case 6: Range Ambiguity in
B-Mode
6.6.1 Background
The conventional B-mode image is formed by
ultrasound beam line, one after another. For each
ultrasound beam line, a very short ultrasound
pulse (< 1 μs) is emitted from the transducer and
then echo signals are received along the beam
line. The next pulse will not be emitted until after
the echo signal from the deepest range of the
FOV from the prior line is received. The time
between successive individual pulse emissions is
called the pulse repetition period (PRP). As a
result, the deeper the range of the FOV, the
longer the PRP is needed.
6.6.2 Findings
Pulse-echo image formation assumes that
returning echo signals are all generated by the
latest pulse emission. As shown in ▶ Fig. 6.17, a
horizontal line formed by reflection echoes from
the bottom of the phantom is mispositioned when
6.6 Case 6: Range Ambiguity in B-Mode
the above mentioned assumption fails. This is
called the range ambiguity artifact.
6.6.3 Discussion
If an echo from a distant structure is received after
the next pulse is transmitted, i.e., beyond the PRP,
the time delay will be counted from the second
pulse emission instead of the first pulse emission.
Consequently, the distance will be mispositioned
to be closer to the transducer than it actually is.
Any scanner parameter setting that shortens the
time interval between pulse emissions, for
example, setting up multiple focal zones, is
susceptible to range ambiguity artifacts.
6.6.4 Resolution
The name “range ambiguity” refers to uncertainties
in the actual range from where the echo signal
occurs. When it occurs, it is likely visible in large
fluid-filled structure, misleading to mimic debris in
the structure. Range ambiguity artifact can be mini-
mized by allowing more time for echoes from deeper
structures to arrive before firing the next pulse.
Fig. 6.17 Range ambiguity artifact
is shown in this phantom image
(pointed by the arrow). The horizontal
line formed by reflection echoes from
the bottom of the phantom is mispo-
sitioned. This occurred when the
reflection echo from the bottom of the
phantom was received after the next
pulse was transmitted; and was thus
counted as echo for the next beam
line.
129
| 09.11.19 - 01:53
Ultrasound Imaging
6.7 Case 7: Shadowing and
Enhancement (Increased
Through Transmission)
6.7.1 Background
As an ultrasound wave propagates through tissue,
its intensity is reduced over distance due to
absorption, scattering, and reflection. The ultra-
sound attenuation coefficient is given in decibels
per centimeter (dB/cm).
6.7.2 Findings
Ultrasound attenuation coefficient is an important
characteristic property of a material. For example,
a fatty infiltrated liver tissue has a higher attenua-
tion coefficient than a healthy liver tissue. For a
focal lesion with higher attenuation coefficient in
comparison to that of its surrounding tissue, shad-
owing occurs (▶ Fig. 6.18). Otherwise, enhance-
ment occurs (▶ Fig. 6.19).
130
6.7.3 Discussion
Reduced echo intensity appears as “shadowing”
distal to a highly attenuating or reflective object
such as a tumor (▶ Fig. 6.18). On the other hand,
increased echo intensity appears as “enhance-
ment” distal to an object with lower attenuation
such as a fluid-filled gallbladder (▶ Fig. 6.19).
6.7.4 Resolution
Shadowing and enhancement are considered
useful image artifacts because these artifacts
indicate the attenuation properties of the object
that causes the occurrence of shadowing or en-
hancement. For example, enhancement is often
used to differentiate cystic structures from solid
structures and shadowing is used to detect
stones, calcified objects, and air. One must be
careful that spatial compounding, an advanced
imaging feature, may affect the appearance of
shadowing or enhancement. When spatial
compounding is activated, each ultrasound beam
is steered into different angles and multiple
Fig. 6.18 Shadowing is shown behind
the breast mass due to higher attenu-
ation in the mass in comparison
to the surrounding tissue.
Fig. 6.19 Enhancement (increased
through transmission) is shown behind
the gallbladder due to lower attenu-
ation in the gallbladder in comparison
to the surrounding tissue.
| 09.11.19 - 01:53
6.7 Case 7: Shadowing and Enhancement (Increased Through Transmission)
steered frames are rendered to create an image
with less speckle and better signal-to-noise
ratio. However, as the spatial compounding
steers the ultrasound beam, the shadowing or
enhancement diverges and loses its intensity,
thus becoming less noticeable. Since spatial
compounding is typically activated in the preset
image protocol, its effect on shadowing and en-
hancement needs to be understood, especially
when the object in question is small. For
example, to see the shadowing associated with a
kidney stone, the operator should deactivate the
spatial compounding feature.
131
| 09.11.19 - 01:53
Ultrasound Imaging
6.8 Case 8: Harmonic Imaging
6.8.1 Background
Harmonic imaging was originally developed on the
basis of nonlinear properties of sound propagation
in ultrasound contrast agents. Later, it was revealed
that the nonlinear effect was also present in tissues.
Tissue harmonic imaging (THI) was developed based
upon the nonlinear effect present in tissue.8
6.8.2 Findings
Harmonic imaging is superior in image quality. A
comparison of harmonic image and conventional
B-mode image is shown in ▶ Fig. 6.20.
6.8.3 Discussion
The term “harmonic” refers to those frequencies that
are integral multiples of the transmitted frequency.
The transmitted frequency is called the fundamental
frequency or the first harmonic frequency. For
example, for a 3.5-MHz transmitted pulse, its funda-
mental frequency, or the first harmonic frequency, is
3.5 MHz. Its second harmonic frequency is twice the
132
fundamental, or in this case, 7 MHz. When a trans-
ducer transmits a burst of ultrasound at a funda-
mental frequency, the sound wave gradually distorts
as it propagates due to the fact that the compres-
sional part of the wave travels slightly faster than
the rarefactional part. Consequently, this distortion
is accompanied by the generation of harmonics that
can be used to form images. This process is called
harmonic imaging.
The benefits of using the harmonic imaging
include improved contrast resolution, reduced
clutter, improved spatial resolution, and reduced
section thickness.8 As shown in ▶ Fig. 6.20, THI has
superior border and tissue definition with reduced
speckles.
6.8.4 Resolution
Many ultrasound imaging protocols have
“harmonic imaging” in the preset as the default
imaging mode instead of conventional B-mode.
This is particularly helpful in imaging “technically
challenging” patients who have thick body walls
or other complicated structures that give rise to
artifacts and clutters.
Fig. 6.20 (a) A comparison of a liver
image acquired in B-mode (left) and
the same patient scanned in tissue
harmonic imaging mode (THI; right) is
shown. Superior image quality is
demonstrated with THI. (b) This panel
shows a comparison of a phantom
image acquired in B-mode (left) and
the same phantom scanned in THI
(right). Mirror artifact (pointed by
arrows) is more prominent in the
B-mode image.
| 09.11.19 - 01:53

6.9 Case 9: Ultrasound Image Display on Scanners and in Reading Rooms

6.9 Case 9: Ultrasound Image
Display on Scanners and in
Reading Rooms
6.9.1 Background
For diagnostic ultrasound imaging, the display
characteristics must be optimal in order to convey
all details and features of ultrasound images to the
human interpretter. In practice, if the operator of
the ultrasound scanner cannot visualize a path-
ology on the scanner display, then the image
cannot be properly acquired, and thus will not be
sent to the picture archiving and communication
system (PACS) and interpreted on reading room
displays. Therefore, an ultrasound scanner display
belongs to the category of diagnostic displays, just
like the reading room displays. Ultrasound scanner
display performance testing is required by the ACR
Ultrasound Accreditation Program.6 In addition,
consistency in image presentation between the
scanner monitor display and the reading room
workstation monitor display must be verified.
6.9.2 Findings
Matching the presentation on the ultrasound
scanner display and the reading room worksta-
tion displays can be challenging as shown in
▶ Fig. 6.21.
6.9.3 Discussion
When a brand new ultrasound scanner is
installed, image presentation consistency
between the scanner monitor display and the
reading room workstation monitor display must
be verified. The ultrasound system, of which the
images are shown in ▶ Fig. 6.21, may have many
different curves to export ultrasound images
from the scanner to PACS. The pixel values of the
image can be altered in order to match the ultra-
sound image presentation on the PACS display.

Fig. 6.21 For an ultrasound scanner, the setting of the system configuration can export the same ultrasound image to
reading room in multiple different ways, resulting different appearances on reading room workstation display as shown
here. LUT, look-up-table.

133
| 09.11.19 - 01:53
Ultrasound Imaging
The limitation of this approach is that it is tail-
ored to a specific characteristic display during
the matching process but not generalizable to
other displays with different characteristic
curves.
What makes establishing presentation consis-
tency between the scanner display and the PACS
display even more challenging is the fact that some
ultrasound manufacturers have created special
display look-up-tables (LUT) on their scanner
display devices to enhance the ultrasound images.
Such image enhancement cannot be replicated
downstream on the PACS display because it is only
available on the scanner display.
134
6.9.4 Resolution
Standardized display systems are needed to map
the presentation values to monitor luminance
regardless of the display monitor or vendor soft-
ware. Such a standardized display system, called
the grayscale standard display function (GSDF),9
has already been developed by the digital
imaging and communication (DICOM). As long as
both the ultrasound scanner monitor and the
reading room monitor follow GSDF and all the
image enhancement is done in the presentation
values, there is a good chance of preserving
presentation consistency.
| 09.11.19 - 01:53
6.10 Case 10: Doppler
Ultrasound Aliasing
6.10.1 Background
The capability to quickly image and quantify blood
flow is unique to diagnostic ultrasound imaging.
The physics principle is straightforward. When-
ever there is a relative motion between a sound
source and a listener, the frequency received by
the listener is shifted from the frequency emitted
by the source. The perceived frequency shift is
called the Doppler shift. We experience the
Doppler shift in routine life. For example, the horn
of a train is shifted to a higher pitch when the train
is approaching the platform and shifted to a lower
pitch when the train is leaving the platform. Math-
ematically, the Doppler shift fD, the difference
between the received sound frequency fr and the
emitted sound frequency fo, is expressed as
following for a blood flow (▶ Fig. 6.22) with
Doppler angle :
2vcosðÞf 0
fD ¼fr  fo ≅
c instrument, the PRF needs to be more than or equal
The Doppler angle  is the angle between the
sound beam and the flow direction. The Doppler
angle affects the detected Doppler shift. When
the Doppler angle is at 90 degrees, no Doppler
shift can be detected. The rule of thumb in clin-
ical applications is to keep the Doppler angle
below 60 degrees.
6.10 Case 10: Doppler Ultrasound Aliasing
6.10.2 Findings
Doppler aliasing in spectral Doppler is shown in
▶ Fig. 6.23. The aliasing is manifested as "wraps
around" from the bottom to the top in spectrum.
Similarly, aliasing can also occur in color
Doppler, showing the reversed color at the peak
flow speed as if the flow direction is reversed
(see ▶ Fig. 6.24).
6.10.3 Discussion
The Doppler shift is detected based upon “sampled”
echo signals. Each time a pulse is emitted and echo
signals are collected along the beam line, the echo
signals are sampled for Doppler shift analysis. There-
fore, the sampling rate of a pulsed Doppler instru-
mentation is equal to the pulse repetition frequency
(PRF). The greater the PRF, the better the rendition
of the Doppler shift signals will be. If the PRF is less
than twice the frequency of the maximum Doppler
signal frequency, then aliasing will occur. The condi-
tion of two times the maximum Doppler shift is
known as the Nyquist criteria. On a pulsed Doppler
to the Nyquist criteria frequency to prevent aliasing
from happening.
6.10.4 Resolution
To eliminate aliasing, the operator can take the
following steps:
Fig. 6.22 Illustration of detecting the
Doppler shift from a flow with a flow
velocity of V, sound speed c, emitting
frequency f0 and receiving frequency fr.
135
| 09.11.19 - 01:53

Ultrasound Imaging

Fig. 6.23 Aliasing on a spectral Doppler display and aliasing elimination steps.

136
| 09.11.19 - 01:53
● Adjust the scale. The PRF is linked to the scale
setting. As the scale increases, the PRF increases.
● Adjust the baseline if the scale has reached its
maximum and the flow is mainly in a single
direction.
● If neither of the above two steps are effective in
eliminating aliasing, lower the transducer
frequency. Lowering the transducer frequency
will lower the Doppler shift, thus lowering the
Nyquist criteria as well.
6.10 Case 10: Doppler Ultrasound Aliasing
Fig. 6.24 Aliasing on a color Doppler
display. The color at the peak speed of
the flow is changed from blue to yellow
as if the flow direction is reversed
(arrow).
● If none of the above steps are successful, increase
the Doppler angle. Increasing the Doppler angle
will lower the Doppler shift, thus lowering the
Nyquist criteria. However, raising the Doppler
angle increases the uncertainty in the accuracy
of Doppler shift measurement and is thus not
desirable.
Having a proper Doppler scale is crucial for good
practices using Doppler ultrasound.
137
| 09.11.19 - 01:53

Ultrasound Imaging

6.11 Review Questions a) Linear array

b) Curved array
6.11.1 Case 1: Pulse-Echo Imaging c) Phased array
Principle and Speed of Sound d) Mechanical sector
Propagation
4. With identical instrumentation settings except
1. If the liver speed of sound is slower than
the frequency selection by the same trans-
1,540 m/s, the size measured by the calipers in
ducer, which image was generated at the
the following image would:
highest frequency?
a) Panel A

a) Be overestimated
b) Be underestimated
c) Be the same

b) Panel B
2. The speed of the sound propagation is
largely determined by which of the following
factor?
a) The transducer frequency
b) The transmission power
c) The medium stiffness
d) The medium attenuation

6.11.2 Case 2: Array Transducers

and Sound Frequency
3. What transducer was used for the
c) Panel C
following image?

138
| 09.11.19 - 01:53

6.11 Review Questions

d) Panel D 6. What QC test should be done to investigate the

cause of the dark streak in the following image
if the streak at the same transducer face loca-
tion appears on many clinical images?

6.11.3 Case 3: Nonuniformity

(Array Transducer Element
Dropouts)
a) Maximum depth of penetration
5. The hypoechoic streak indicated by the arrow b) Distance accuracy
in the following image is most likely caused c) Image uniformity
by? d) Low contrast detectability

6.11.4 Case 4: Pulse-Echo Imaging

Acquisition Controls
7. What acquisition control should be increased if
more depth of penetration is needed?
a) Transducer frequency
b) Dynamic range
c) Overall gain
d) Transmit power

8. What acquisition control should be adjusted if

a horizontal band of darkness is seen?
a) Improper gain setting a) TGC
b) Compound imaging b) Dynamic range
c) Dead transducer elements c) Overall gain
d) Inappropriate transducer pressure d) Transmit power

139
| 09.11.19 - 01:53

Ultrasound Imaging

6.11.5 Case 5: Reflection c) Reduce transmit power

d) Reduce transducer frequency
(Boundary Conditions)—
Reverberation Artifacts
6.11.7 Case 7: Shadowing and
9. What is the name of the artifact indicated by
Enhancement (Increased Through
the following image?
Transmission)
13. What is the name of the artifact indicated by
the arrow in the following image?

a) Enhancement
b) Comet tail
c) Beam width
d) Range ambiguity

10. What is the cause for the artifact indicated by

the arrow the image from the previous
question? a) Shadowing
a) Multiple refractions b) Comet-tail
b) Multiple reflections c) Beam width
c) Beam steering d) Ring-down
d) Beam diverging
14. How is the attenuation of the thyroid nodule
compared with the attenuation of the
6.11.6 Case 6: Range Ambiguity in
B-Mode
11. Which of the following is most likely to occur
when pulse repetition period is too short and
an echo signal from the prior beam line is
mispositioned in the current beam line?
a) Speed artifact
b) Aliasing
c) Mirror artifact
d) Range ambiguity
surrounding tissue in the following image?
a) The nodule is more attenuating
12. What can be done to minimize range
b) The nodule is less attenuating
ambiguity artifacts?
c) the same
a) Reduce overall gain
b) Reduce number of focal zones

140
| 09.11.19 - 01:53

6.11 Review Questions

6.11.8 Case 8: Harmonic Imaging b) Scanner display performance

c) Primary interpretation display performance
15. Which of the following is a benefit using the
d) Contrast resolution
THI instead of conventional B-mode imaging?
a) Improved depth of penetration
b) Reduced clutter
6.11.10 Case 10: Doppler
c) Increased temporal resolution Ultrasound Aliasing
d) Reduced sound frequency 19. What is of the artifact indicated by the arrow in
the following image?
16. Which ultrasound imaging technology
was effective to improve image quality and
reduce artifact in abdominal ultrasound
imaging of a patient with a very thick body
wall?
a) Spatial compounding
b) Ultrafast imaging
c) Harmonic imaging
d) Coded excitation

6.11.9 Case 9: Ultrasound Image

Display on Scanners and in
Reading Rooms
a) Aliasing
17. Advantages of enabling the GSDF on ultrasound b) Spectral broadening
scanner display include all of the following c) Twinkle
except: d) Flash
a) Image enhancement performed by the ultra-
sound scanner display 20. What can be done to minimize aliasing?
b) No custom matching needed on the reading a) Reduce wall filter
room workstation displays b) Reduce gain
c) Presentation consistency among various c) Increase scale
displays in the imaging chain d) Increase frequency
d) More streamlined performance evaluation
and quality control
References
18. Which of the following is a required QC test by [1] Zagzebski JA. Pulse-echo ultrasound instrumentation.
Essentials of Ultrasound Physics. St. Louis: Mosby,
the ACR Ultrasound Accreditation Program? 1996:46–47
a) Distance accuracy

141
| 09.11.19 - 01:53
Ultrasound Imaging
[2] Zagzebski JA. Physics of diagnostic ultrasound. Essentials of
Ultrasound Physics. St. Louis: Mosby, 1996:6
[3] Dudley NJ, Gibson NM, Fleckney MJ, Clark PD. The effect of
speed of sound in ultrasound test objects on lateral resolu-
tion. Ultrasound Med Biol. 2002; 28(11–12):1561–1564
[4] Hangiandreou NJ, Stekel SF, Tradup DJ, Gorny KR, King DM.
Four-year experience with a clinical ultrasound quality con-
trol program. Ultrasound Med Biol. 2011; 37(8):1350–1357
[5] Powis RL, Moore GW. The silent revolution: catching up with
the contemporary composite transducer. J Diagn Med Sonogr.
2004; 20:395–405
[6] American College of Radiology (ACR). Ultrasound accredita-
tion program requirements: http://www.acraccreditation.
142
org/~/media/ACRAccreditation/Documents/Ultrasound/
Requirements.pdf?la=en (accessed 12/26/2018)
[7] Avruch L, Cooperberg PL. The ring-down artifact. J Ultrasound
Med. 1985; 4(1):21–28
[8] Anvari A, Forsberg F, Samir AE. A primer on the physical prin-
ciples of tissue harmonic imaging. Radiographics. 2015; 35
(7):1955–1964
[9] DICOM PS3. 14 2018e – Grayscale Standard Display Function,
NEMA. http://dicom.nema.org/medical/dicom/current/out-
put/html/part14.html (accessed 12/31/2018)
| 09.11.19 - 01:53

7 Image Processing
Jonathon A. Nye and Randahl C. Palmer

Introduction interpretation. In addition, the combination of infor-

mation from two or more modalities can increase
Image processing and display are critical components both sensitivity and specificity compared to a single
in the imaging workflow chain. They have tremen- exam. This chapter presents basic image processing
dous diagnostic utility for disease detectability and concepts that are used in medical imaging.

143
| 09.11.19 - 01:53
Image Processing
7.1 Case 1: Filtering and Edge
Enhancement
7.1.1 Background
Presentation of an anteroposterior (AP) chest
radiograph.
7.1.2 Findings
● Presentation of post image filtering by convolu-
tion with kernels designed to extract features of
different frequencies.
● Kernels are commonly used to lower noise or
enhance edge information.
7.1.3 Discussion
Filtering is one of the most basic imaging proc-
essing steps to enhance image contrast. It can be
applied either in frequency space through use of
the Fourier transform or in image space through
use of the convolution process.1 In image space,
filtering involves construction of a kernel that is
moved across the image. All pixels within the
kernel are averaged and that average is placed in
Fig. 7.1 1D example of a convolution of an edge (open circles) with a Gaussian kernel. Data are rounded to the nearest
whole number for easier display. The plot shows the original data including the same information following a Gaussian
smoothing operation (solid line) and a hybrid Gaussian–Laplacian edge enhancement operation (dashed line).
144
a new image. A common kernel is a Gaussian
function, which is a low-pass operation that
reduces noise (e.g., quantum mottle) and
improves visibility of low-contrast features.
High-contrast features can be enhanced by
adding negative lobes to the kernel. This process
improves high-contrast resolution but also
increases noise since both of these features are
high-frequency components of an image. Kernels
are normalized to preserve the scale of the orig-
inal image. The convolution process is described
in ▶ Fig. 7.11 where discrete kernel values are
multiplied by the pixel values that fall under-
neath the function and the sum of these products
is placed in a new image. The kernel is then
shifted and the process is repeated. ▶ Fig. 7.2
demonstrates the change in contrast of a chest
radiograph following Gaussian smoothing and
Gaussian–Laplacian edge enhancement.
7.1.4 Resolution
Convolution using kernels designed to lower
noise or enhance edge information alters
contrast and can improve the detectability of
anatomical features such as soft-tissue masses or
bone fractures.
| 09.11.19 - 01:53

7.1 Case 1: Filtering and Edge Enhancement

Fig. 7.2 (a) Chest X-ray without application of a postprocessing convolution filter. (b) Convolved with a 10-pixel full-
width-at-half-maximum Gaussian filter. (c) Convolved with a 10-pixel full-width-at-half-maximum Gaussian–Laplacian
edge-enhancement filter.

145
| 09.11.19 - 01:53
Image Processing
7.2 Case 2: Maximum Intensity
Projection
7.2.1 Background
● Whole-body [18F] fluorodeoxyglucose ([18F]
FDG) positron emission tomography (PET) of a
patient with extensive disease.
● The image volume is processed with a ray-
tracing technique called maximum intensity
projection (MIP), to highlight hyper-metabolic
activity throughout the volume (▶ Fig. 7.3).
7.2.2 Findings
● Compared to a standard coronal slice, the MIP
image permits visualization at depth within a 3D
volume on a 2D display.
● The MIP volume can be rotated to improve visu-
alization of hot lesions that lie along the same
ray path but in different planes.
7.2.3 Discussion
MIP is a simple but powerful processing tool
used to visualize a 3D volumetric dataset. The
most straightforward implementation of the
method involves using a ray-tracing technique
that projects the maximum pixel value along
parallel paths within a 3D volume on to a 2D
146
viewing plane (▶ Fig. 7.4). The volume can be
oriented in any direction or rotated incremen-
tally after each ray tracing to produce many 2D
projections that can be played back as a movie.
The MIP process gives the perception of looking
through a volume as opposed to the conven-
tional methods of paging through slices. It is
used extensively in nuclear medicine tomog-
raphy (e.g., PET/SPECT) to quickly identify hot
spots in a large 3D volume. Two hot areas of
interest that lie along the same ray path can
mask one another, therefore rotating the volume
can reveal these superimposed areas.
7.2.4 Resolution
MIP is a simple yet powerful visual rendering
method that gives information about an entire 3D
volume without the need to scroll through slices.
The method is completely automated, commonly
displayed in the coronal orientation, and may be
oriented in any direction. More sophisticated
implementations can be performed where pixels
along the visualization vector are weighted by or
limited to a certain depth within the 3D volume.
This process is used, in combination with other
processing techniques, to construct synthetic 2D
mammograms from breast tomosynthesis data.
Other common uses of MIP include visualization of
vascular structures in computed tomography
angiography and magnetic resonance (MR) time-
of-flight.2,3,4
Fig. 7.3 (a) Coronal slice of a [18F] FDG
whole-body PET/CT showing extensive
disease. (b) Maximum intensity
projection processed volume in the
coronal orientation along the anterior-
posterior direction. Note the visual-
ization of lesions at deeper slices that
are out of plane in (a).
| 09.11.19 - 01:53

7.2 Case 2: Maximum Intensity Projection

Fig. 7.4 (a) Illustration of the

maximum intensity projection (MIP)
process of ray tracing a 3D volume to
create a 2D MIP image. (b) A ray path
through pixels finds the maximum
value and displays that value on the
workstation.

147
| 09.11.19 - 01:53
Image Processing
7.3 Case 3: Fused Image Display
of Multiple Modalities
7.3.1 Background
● Independently collected MR and PET images
were acquired, registered, and displayed as a
single image (▶ Fig. 7.5).
● The T2-weighted fluid-attenuation inversion
recovery (FLAIR) MRI, presented in grayscale,
provides structural details that can be visually
correlated with functional metabolic informa-
tion provided by the [18F] FDG brain PET
image.
7.3.2 Findings
● Limited or poor anatomical detail in PET can be
augmented by coregistering and fusing these
data with a high-resolution anatomical image.
Fig. 7.5 An example of the alpha blending technique using an magnetic resonance image in grayscale and positron
emission tomography image in rainbow color.
148
● Data fusion allows for better identification of
anatomical boundaries of disease and visual
correlation with changes in metabolism.
7.3.3 Discussion
Display fusion, or color blending, of two or more
images is a widely employed technique for
displaying and interpreting functional imaging
data.5,6 Commonly, the underlying image is a
structural modality (e.g., MRI, CT) displayed in
grayscale and the overlying image is a functional
modality (e.g., PET, SPECT, MRS) displayed in a
false color scale. The fusion process is commonly
called alpha blending, where images are
converted to a 24-bit color image (e.g., red,
green, and blue channels) and a transparency
value (alpha) is assigned to the blended image.
Monitors shipped with common desktop
computers are 24-bit, having 3 channels each
with 256 shades of color. The displayed image is
then a combination of two color scales, grayscale
| 09.11.19 - 01:53
7.3 Case 3: Fused Image Display of Multiple Modalities
and false color, that gives the perception of
transparency depending on the choice of the
alpha value.
7.3.4 Resolution
The example in ▶ Fig. 7.5 demonstrates the alpha
blending image fusion technique using a linear
combination of display scales from two images.
More sophisticated blending techniques are avail-
able that can preserve certain features, for
example, thresholding the PET image to display
only standardized uptake values above a prede-
fined value. Image fusion has enabled hybrid
imaging to grow into a powerful diagnostic tool
increasing both sensitivity and specificity of inter-
pretation compared to viewing one of the two
modalities alone.7
149
| 09.11.19 - 01:53

Image Processing

7.4 Case 4: Multimodality ● Spatial coregistration can be accomplished using

Image Registration
7.4.1 Background
● [18F] FDG and T2-weighted FLAIR brain images
were acquired in a patient with epilepsy.
● Images were collected on different instruments,
therefore there is a need to spatially align them
to the same reference orientation.
7.4.2 Findings
● Fused display of the images show misalignment
because these data were acquired on stand-
alone systems (▶ Fig. 7.6a).
automated routines based on similarity metrics
including entropy and mutual information.
7.4.3 Background
Patient data collected from different scanners can
be coregistered to correlate voxel intensities of
function (e.g., PET, SPECT) with areas of
morphology (CT, MRI). For imaging data collected
on stand-alone systems (nonhybrid), a software
solution is needed. There are a large number of
coregistration similarity metrics for determining
whether two images are aligned well. Two
common approaches include use of external
surface landmarks and correlation of voxel inten-
sity but these methods are either cumbersome to

Fig. 7.6 Examples of misregistered (a, c) and registered (b, d) images of positron emission tomography with magnetic
resonance imaging (a, b) and computed tomography (c, d). The second column are the joint histograms of pixel
intensities between the respective images. The third column lists the calculated joint entropy and mutual information of
the images in their orientation as displayed in the first column. Note that lower joint entropy and higher mutual
information indicate better spatial alignment.

150
| 09.11.19 - 01:53
implement or present difficulties when voxels of
two different modalities are weakly correlated,
respectively. A popular similarity metric for inter-
modality image registration is the one based on
entropy.8,9 Entropy is a measure of dispersion in
the probability distribution. The probably distribu-
tion between two images can be represented by
plotting a 2D histogram. The joint histogram
counts how many times a grayscale value corre-
spondence occurs; the brighter the pixel in the
joint histogram, the larger the correspondence fre-
quency. Entropy can be thought of as a measure of
the histogram dispersion. Minimizing the disper-
sion of the joint histogram occurs when the joint
entropy between two image modalities is mini-
mized. A common implementation of entropy is
maximization of mutual information (MI) given as
Entropy ðMRIÞ þ Entropy ðPETÞ − Entropy ðMRI þ PETÞ
MI is the most commonly used similarity metric
and is considered highly robust and reliable.
▶ Fig. 7.6 demonstrates the differences in sensi-
tivity between entropy and MI when comparing
misregistered and registered PET images with
either CT or MRI.
7.4.4 Resolution
Image registration is an optimization problem
where the goal is to match corresponding features
7.4 Case 4: Multimodality Image Registration
by maximizing a similarity measure. Similarity
measures based on entropy have prevailed, as they
are highly reliable when aligning two images from
different modalities.
References
[1] Brody WR. In: Bankman IN, ed. Handbook of Medical Image
Processing and Analysis. 2nd ed. Burlington: Academic Press;
2009
[2] Prokop M, Shin HO, Schanz A, Schaefer-Prokop CM. Use of
maximum intensity projections in CT angiography: a basic
review. Radiographics. 1997; 17(2):433–451
[3] Saloner D. The AAPM/RSNA physics tutorial for residents. An
introduction to MR angiography. Radiographics. 1995; 15(2):
453–465
[4] Hofman MS, Hicks RJ. How we read oncologic FDG PET/CT.
Cancer Imaging. 2016; 16(1):35
[5] Bican J, Janeba D, Táborská K, Veselý J. Image overlay using
alpha-blending technique. Nucl Med Rev Cent East Eur. 2002;
5(1):53
[6] Watson CC, Townsend DW, Bendriem B. PET/CT systems. In:
Wernick MN, Aarsvold JN, eds. Emission Tomography. San
Diego: Academic Press; 2004:195–212
[7] Hany TF, Steinert HC, Goerres GW, Buck A, von Schulthess
GK. PET diagnostic accuracy: improvement with in-line PET-
CT system: initial results. Radiology. 2002; 225(2):575–581
[8] Viergever MA, Maintz JBA, Klein S, Murphy K, Staring M,
Pluim JPW. A survey of medical image registration—under
review. Med Image Anal. 2016; 33:140–144
[9] Oliveira FP, Tavares JM. Medical image registration: a review.
Comput Methods Biomech Biomed Engin. 2014; 17(2):73–93
151
| 09.11.19 - 01:53

Answer Key
Chapter 1 Fluoroscopy Chapter 3 Computed
1. a Tomography
2. b 1. a
3. e 2. b
4. c 3. a
5. b 4. a
6. a 5. d
7. d 6. b
8. a 7. c
9. a 8. d
10. b
9. b
11. a
10. a
12. c
11. a
13. b
12. b
14. c
13. b
15. c
14. b
16. b
15. a
17. b
16. a
18. a
17. d
19. d
18. a
20. c
19. c
20. b
Chapter 2 Mammography
1. a Chapter 4 Magnetic Resonance
2. b
Imaging
3. c
4. b 1. b
5. b 2. c
6. b 3. b
7. b 4. b
Due to the limited angle scan, depth resolu- 5. b
tion in digital tomosynthesis is much lower 6. a
than the in-plane resolution. 7. c
8. c 8. c
9. d 9. d
10. b 10. c
11. b 11. a
12. b 12. a
13. e 13. b
14. e 14. c
15. b 15. b
16. d 16. c
17. b 17. d
18. d 18. b
19. c 19. d
20. d
21. c
22. b

152
| 09.11.19 - 01:53

Answer Key

Chapter 5 Nuclear Medicine [Zagzebski JA. Physics of diagnostic ultra-

sound. Essentials of Ultrasound Physics. St.
1. b Louis: Mosby; 1996:26-27]
2. a 5. c
3. a The nonuniformity near surface is typically
4. b due to dead transducer elements.
5. a [ACR technical standard for diagnostic
6. c medical physics performance monitoring of
real time ultrasound equipment, Revised
7. b
2016]
8. b
6. c
9. d
Image uniformity is an effective phantom
10. a
test for revealing transducer element drop-
11. a
outs.
12. c
[Hangiandreou NJ, Stekel SF, Tradup DJ,
13. b
Gorny KR, King DM. Four-year experience
14. d
with a clinical ultrasound quality control
15. b
program. Ultrasound Med Biol. 2011; 37
16. a
(8):1350-1357]
17. a
7. d
18. b
Only the increase of the transmit power
19. c
will increase the depth of penetration. The
20. b
increase of the dynamic range or the
overall gain will not affect the maximum
Chapter 6 Ultrasound Imaging depth of penetration. Increasing the trans-
ducer frequency will reduce penetration.
1. a [Zagzebski JA. Pulse-echo ultrasound
If the true speed of sound propagation is instrumentation. Essentials of Ultrasound
slower than 1,540 m/s, it takes longer time Physics. St. Louis: Mosby; 1996]
to travel; thus the distance will be overesti- 8. a
mated. The horizontal band of darkness may be
[Zagzebski JA. Pulse-echo ultrasound caused by insufficient amplification at the
instrumentation. Essentials of Ultrasound specific depth and can be adjusted by
Physics. St. Louis: Mosby, 1996:46-47] sliding the TGC knob that controls the
2. c amplification at the corresponding depth.
The speed of sound is dependent on the [Zagzebski JA. Pulse-echo ultrasound
properties of the medium such as its instrumentation. Essentials of Ultrasound
density and compressibility. Physics. St. Louis: Mosby; 1996]
[Zagzebski JA. Physics of diagnostic ultra- 9. b
sound. Essentials of Ultrasound Physics. St. [Zagzebski JA. Pulse-echo ultrasound
Louis: Mosby, 1996:6] instrumentation. Essentials of Ultrasound
3. b Physics. St. Louis: Mosby; 1996]
Curved array produces acoustic beams 10. b
diverging as the depth increases and its [Zagzebski JA. Pulse-echo ultrasound
transducer surface has a wide footprint. instrumentation. Essentials of Ultrasound
[Zagzebski JA. Physics of diagnostic ultra- Physics. St. Louis: Mosby; 1996]
sound. Essentials of Ultrasound Physics. St. 11. d
Louis: Mosby; 1996:34-3] [Hangiandreou NJ, O'Brien RT, Zagzebski JA,
4. a Delaney FA. Ultrasound corner: range
This image has the lowest depth of penetra- ambiguity artifact. Vet Radiol Ultrasound.
tion due to high attenuation at high fre- 2001; 42(6):542-545]
quency. 12. b

153
| 09.11.19 - 01:53

Answer Key

[Hangiandreou NJ, O'Brien RT, Zagzebski JA, 17. a

Delaney FA. Ultrasound corner: range [On-Line Report No AAPM. 03, Assessment
ambiguity artifact. Vet Radiol Ultrasound. of Display Performance for Medical Imaging
2001; 42(6):542-545] Systems, American Association of Physicists
13. a in Medicine, 2005. https://www.aapm.org/
[Zagzebski JA. Pulse-echo ultrasound pubs/reports/OR_03.pdf (accessed on 12/
instrumentation. Essentials of Ultrasound 31/2018)]
Physics. St. Louis: Mosby; 1996] 18. b
14. b [American College of Radiology (ACR).
[Zagzebski JA. Pulse-echo ultrasound Ultrasound accreditation program require-
instrumentation. Essentials of Ultrasound ments: http://www.acraccreditation.org/
Physics. St. Louis: Mosby; 1996] ~/media/ACRAccreditation/Documents/
15. b Ultrasound/Requirements.pdf?la=en
[Anvari A, Forsberg F, Samir AE. A primer (accessed 12/26/2018)]
on the physical principles of tissue 19. a
harmonic imaging. Radiographics. 2015; 35 [Zagzebski JA. Pulse-echo ultrasound
(7):1955-1964] instrumentation. Essentials of Ultrasound
16. c Physics. St. Louis: Mosby; 1996]
[Anvari A, Forsberg F, Samir AE. A primer 20. c
on the physical principles of tissue [Zagzebski JA. Pulse-echo ultrasound
harmonic imaging. Radiographics. 2015; 35 instrumentation. Essentials of Ultrasound
(7):1955-1964] Physics. St. Louis: Mosby; 1996]

154
| 09.11.19 - 01:53

Index
A
abnormal dark fluid seen in bladder of an – displayed volume CT dose index CT fluoroscopy
axial single-Shot T2-weighted sequence –– background 51 – background 12
– background 76 –– discussion 51 – discussion 12
– discussion 76 –– findings 51 – findings 12
– findings 76 –– resolution 51 – resolution 12
– resolution 78 – image quality variation with recon-
antiscatter grids structed slice thickness
– background 8 –– background 47 D
– discussion 8 –– discussion 47
–– findings 47 dark etching appears at boundary of fat
– findings 8
–– resolution 47 and soft-tissue layers
– resolution 9
– image quality variation with recon- – background 69
array transducers and sound frequency
struction filter – discussion 69
– background 120
–– background 49 – findings 69
– discussion 121
–– discussion 49 – resolution 69
– findings 120
–– findings 49 degraded image quality from improper
artifact due to detector row dropout
–– resolution 49 collimator
– background 35
– kV selection on image quality and dose, – background 112
– discussion 35
effect of – discussion 112
– findings 35
–– background 46 – findings 112
– resolution 35
–– discussion 46 – resolution 112
artifact due to imperfection in compression
–– findings 46 degraded resolution of whole-body planar
paddle
99mTc methylene diphosphonate image
– background 37 –– resolution 46
– metal artifact – background 91
– discussion 37
–– background 54 – discussion 91
– findings 37
–– discussion 54 – findings 91
– resolution 37
–– findings 54 – resolution 91
attenuation correction in PET
–– resolution 54 digital breast tomosynthesis
– background 99
– motion artifact – background 30
– discussion 99
–– background 55 – discussion 30
– findings 99
–– discussion 55 – findings 31
– resolution 100
–– findings 55 – resolution 31
–– resolution 55 digital subtraction angiography and mo-
B – partial volume artifact tion artifacts
–– background 53 – background 14
beam hardening artifact – discussion 14
–– discussion 53
– background 52 – findings 14
–– findings 53
– discussion 52 – resolution 15
–– resolution 53
– findings 52 discrete image ghosts on abdominal imag-
– patient size on CT number accuracy, ef-
– resolution 52 ing, appearance of
fect of
bremsstrahlung imaging of 90Y micro- – background 60
–– background 45
spheres liver embolization – discussion 60
–– discussion 45
– background 110 – findings 60
–– findings 45
– discussion 110 – resolution 61
–– resolution 45
– findings 110 displayed volume CT dose index
– ring artifact
– resolution 110 – background 51
–– background 44
–– discussion 44 – discussion 51
–– findings 44 – findings 51
C – resolution 51
–– resolution 44
collimation cone beam computed tomography doppler ultrasound aliasing
– background 6 – background 20 – background 135
– discussion 6 – discussion 20 – discussion 135
– findings 6 – findings 20 – findings 135
– resolution 6 – resolution 20 – resolution 135
computed tomography correct acquisition image matrix size
– beam hardening artifact – background 106
–– background 52 – discussion 106
E
–– discussion 52 – findings 106 EMI artifact due to LVAD device
–– findings 52 – resolution 107 – background 39
–– resolution 52 – discussion 39

155
| 09.11.19 - 01:53

Index

– findings 39 –– background 10 –– resolution 146

– resolution 39 –– discussion 10 – multimodality image registration
equalization filters –– findings 10 –– background 150
– background 18 –– resolution 11 –– discussion 150
– discussion 18 – reference air kerma and skin dose –– findings 150
– findings 18 –– background 4 –– resolution 151
– resolution 18 –– discussion 4 image quality variation with reconstructed
extra field-of-view anatomy on inferior –– findings 4 slice thickness
portion of sagittal 3D T2-weighted ac- –– resolution 5 – background 47
quisition of spine, appearance of – SID, ABC, and radiation output 12 – discussion 47
– background 65 –– background 2 – findings 47
– discussion 65 –– discussion 2 – resolution 47
– finding 65 –– findings 2 image quality variation with reconstruc-
– resolution 66 –– resolution 2 tion filter
focal spot size selection in magnification – background 49
views – discussion 49
F – background 27 – findings 49
– discussion 27 – resolution 49
fat-suppressed sequences, application of
– findings 27 image smoothing
– background 71
– resolution 28 – background 104
– discussion 71
fused image display of multiple modalities – discussion 104
– findings 71
– background 148 – findings 104
– resolution 72
– discussion 148 – resolution 105
filtering and edge enhancement
– findings 148 iterative reconstruction and choosing the
– background 144
– resolution 149 number of iterations and subsets
– discussion 144
– background 101
– findings 144
– discussion 101
– resolution 144
fluoroscopy
H – findings 101
– resolution 101
– antiscatter grids harmonic imaging
–– background 8 – background 132
–– discussion 8 – discussion 132
–– findings 8 – findings 132
K
–– resolution 8 – resolution 132 kV selection on image quality and dose,
– collimation hyperintensity appears bilaterally at the effect of
–– background 6 level of the internal auditory canal on – background 46
–– discussion 6 diffusion-weighted MRI, affecting visu- – discussion 46
–– findings 6 alization of surrounding structures – findings 46
–– resolution 7 – background 62 – resolution 46
– cone beam computed tomography – discussion 62
–– background 20 – findings 62
–– discussion 20 – resolution 63 M
–– findings 20
magnetic resonance imaging
–– resolution 21
– abnormal dark fluid seen in bladder of
– CT fluoroscopy I an axial single-Shot T2-weighted
–– background 12
image post-processing on appearance sequence, but not on location-matched
–– discussion 12
– background 32 3D T2 acquistion
–– findings 12
– discussion 32 –– background 76
–– resolution 12
– findings 34 –– discussion 76
– digital subtraction angiography and
– resolution 34 –– findings 76
motion artifacts
image processing –– resolution 78
–– background 14
– filtering and edge enhancement – dark etching appears at boundary of fat
–– discussion 14
–– background 144 and soft-tissue layers 67
–– findings 14
–– discussion 144 –– background 69
–– resolution 15
–– findings 144 –– discussion 69
– equalization filters
–– resolution 144 –– findings 69
–– background 18
– fused image display of multiple modal- –– resolution 69
–– discussion 18
ities – appearance of discrete image ghosts on
–– findings 18
–– background 148 abdominal imaging 60
–– resolution 18
–– discussion 148 –– background 60
– modes and dose
–– findings 148 –– discussion 60
–– background 16
–– resolution 149 –– findings 60
–– discussion 16
– maximum intensity projection –– resolution 61
–– findings 16
–– background 146 – extra field-of-view anatomy on inferior
–– resolution 16
–– discussion 146 portion of sagittal 3D T2-weighted ac-
– patient shielding
–– findings 146 quisition of spine, appearance of 65

156
| 09.11.19 - 01:53

Index

–– background 65 –– resolution 39
–– discussion 65 – equations 42
N
–– finding 65 – focal spot size selection in magnifica- nonuniformity (array transducer element
–– resolution 66 tion views 27 dropouts)
– fat-suppressed sequences, application –– background 27 – background 122
of 71 –– discussion 27 – discussion 122
–– background 71 –– findings 27 – findings 122
–– discussion 71 –– resolution 28 – resolution 122
–– findings 71 – image post-processing on appear- nuclear medicine
–– resolution 72 ance 32 – attenuation correction in PET
– hyperintensity appears bilaterally at –– background 32 –– background 99
level of internal auditory canal on –– discussion 32 –– discussion 99
diffusion-weighted MRI, affecting visu- –– findings 34 –– findings 99
alization of surrounding structures –– resolution 34 –– resolution 99
–– background 62 – magnification imaging – bremsstrahlung imaging of 90Y micro-
–– discussion 62 –– background 26 spheres liver embolization
–– findings 62 –– discussion 26 –– background 110
–– resolution 63 –– findings 26 –– discussion 110
– post contrast T1-weighted gradient –– resolution 26 –– findings 110
echo reveals patchy enhancement in an- – microcalcification-like appearance –– resolution 110
terior septal wall 79 caused by detector artifact – correct acquisition image matrix size
–– background 79 –– background 36 –– background 106
–– discussion 79 –– discussion 36 –– discussion 106
–– findings 79 –– findings 36 –– findings 106
–– resolution 81 –– resolution 36 –– resolution 107
– precontrast, axial 3D T1-weighted gra- – patient motion causing blurred paren- – degraded image quality from improper
dient echo with fat suppressions 67 chymal structure collimator
–– background 67 –– background 38 –– background 112
–– discussion 67 –– discussion 38 –– discussion 112
–– findings 67 –– findings 38 –– findings 112
–– resolution 68 –– resolution 38 –– resolution 112
– signal-to-noise variation across FOV, – x-ray acquisition technique factors – degraded resolution of whole-body pla-
creating nondiagnostic image quality –– background 29 nar 99mTc methylene diphosphonate
–– background 82 –– discussion 29 image
–– discussion 82 –– findings 29 –– background 91
–– findings 82 –– resolution 30 –– discussion 91
–– resolution 84 maximum intensity projection –– findings 91
– T1-weighted gradient echo of abdomen – background 146 –– resolution 92
shows marked artifact medially on both – discussion 146 – image smoothing
coronal and axial FOV, obscuring visual- – findings 146 –– background 104
ization of soft tissues – resolution 146 –– discussion 104
–– discussion 74 metal artifact –– findings 104
–– Fbackground 74 – background 54 –– resolution 105
–– findings 74 – discussion 54 – iterative reconstruction and choosing
–– resolution 75 – findings 54 the number of iterations and subsets
magnification imaging – resolution 54 –– background 101
– background 26 microcalcification-like appearance caused –– discussion 101
– discussion 26 by detector artifact –– findings 101
– findings 26 – background 36 –– resolution 101
– resolution 26 – discussion 36 – patient motion in myocardial perfusion
mammography – findings 36 imaging
– artifact due to detector row dropout 35 – resolution 36 –– background 108
–– background 35 modes and dose –– discussion 108
–– discussion 35 – background 16 –– findings 108
–– findings 35 – discussion 16 –– resolution 108
–– resolution 35 – findings 16 – positron range on image quality and
– artifact due to imperfection in compres- – resolution 16 resolution
sion paddle 37 motion artifact –– background 94
–– background 37 – background 55 –– findings 94
–– discussion 37 – discussion 55 –– discussion 94
–– findings 37 – findings 55 –– resolution 96
–– resolution 37 – resolution 55 – standardized uptake value in positron
– EMI artifact due to LVAD device multimodality image registration emission tomography
–– background 39 – background 150 –– background 97
–– discussion 39 – findings 150 –– discussion 97
–– findings 39 – resolution 151

157
| 09.11.19 - 01:53

Index

–– findings 97 – findings 133

–– resolution 98
R – discussion 133
range ambiguity in B-mode – resolution 134
– background 129 ultrasound imaging
P – findings 129 – array transducers and sound frequency
– discussion 129 –– background 120
partial volume artifact
– resolution 129 –– discussion 121
– background 53
reference air kerma and skin dose –– findings 120
– discussion 53
– background 4 –– resolution 121
– findings 53
– discussion 4 – doppler ultrasound aliasing
– resolution 53
– findings 4 –– background 135
patient motion causing blurred parenchy-
– resolution 5 –– discussion 135
mal structure
reflection (boundary conditions)-reverber- –– findings 135
– background 38
ation artifacts –– resolution 135
– discussion 38
– background 127 – harmonic imaging
– findings 38
– discussion 127 –– background 132
– resolution 38
– findings 127 –– findings 132
patient motion in myocardial perfusion
– resolution 128 –– discussion 132
imaging
ring artifact –– resolution 132
– background 108
– background 44 – nonuniformity (array transducer ele-
– discussion 108
– discussion 44 ment dropouts)
– findings 108
– findings 44 –– background 122
– resolution 108
– resolution 44 –– discussion 122
patient shielding
–– findings 122
– background 10
–– resolution 123
– discussion 10
– findings 10
S – pulse-echo imaging acquisition
–– background 124
– resolution 11 shadowing and enhancement
–– findings 124
patient size on CT number accuracy, effect – background 130
–– discussion 124
of – findings 130
–– resolution 126
– background 45 – discussion 130
– pulse-echo imaging principle and speed
– discussion 45 – resolution 130
of sound propagation
– findings 45 SID, ABC, and radiation output
–– background 118
– resolution 45 – background 2
–– discussion 118
positron range on image quality and reso- – discussion 2
–– findings 118
lution – findings 2
–– resolution 118
– background 94 – resolution 2
– range ambiguity in B-mode
– findings 94 signal-to-noise variation across FOV, creat-
–– background 129
– discussion 94 ing nondiagnostic image quality
–– findings 129
– resolution 96 – background 82
–– discussion 129
post contrast T1-weighted gradient echo – discussion 82
–– resolution 129
reveals patchy enhancement in anterior – findings 82
– reflection (boundary conditions)-rever-
septal wall – resolution 84
beration artifacts
– background 79 standardized uptake value in positron
–– background 127
– discussion 79 emission tomography
–– discussion 127
– findings 79 – background 97
–– findings 127
– resolution 81 – discussion 97
–– resolution 128
precontrast, axial 3D T1-weighted gradient – findings 97
– shadowing and enhancement
echo with fat suppressions – resolution 98
–– background 130
– background 67
–– findings 130
– discussion 67
– findings 67 T –– discussion 130
–– resolution 130
– resolution 68
T1-weighted gradient echo of abdomen – ultrasound image display on scanners
pulse-echo imaging acquisition
shows marked artifact medially on both and in reading rooms
– background 124
coronal and axial FOV, obscuring visual- –– background 133
– findings 124
ization of soft tissues –– findings 133
– discussion 124
– background 74 –– discussion 133
– resolution 126
– discussion 74 –– resolution 134
pulse-echo imaging principle and speed of
– findings 74
sound propagation
– resolution 75
– background 118
– discussion 118
X
– findings 118 U X -ray acquisition technique factors
– resolution 118 – background 29
ultrasound image display on scanners and – discussion 29
in reading rooms – findings 29
– background 133 – resolution 30

158