CME
JOURNAL OF MAGNETIC RESONANCE IMAGING 34:1212–1217 (2011)
Technical Note
Supine Breast MRI
Peter Siegler, PhD,1* Claire M.B. Holloway, MD, PhD,2 Petrina Causer, MD,3
Gilbert Thevathasan, BEng,4 and Donald B. Plewes, PhD1
Purpose: To achieve high-quality unilateral supine
breast magnetic resonance imaging (MRI) as a step to
facilitate image aiding of clinical applications, which
are often performed in the supine position. Contrast-
enhanced breast MRI is a powerful tool for the diagno-
sis of cancer. However, prone patient positioning typi-
cally used for breast MRI hinders its use for image
aiding.
Materials and Methods: A fixture and a flexible four-ele-
ment receive coil were designed for patient-specific shap-
ing and placement of the coil in close conformity to the
supine breast. A 3D spoiled gradient sequence was modi-
fied to incorporate compensation of respiratory motion.
The entire setup was tested in volunteer experiments and
in a pilot patient study.
Results: The flexible coil design and the motion compen-
sation produced supine breast MR images of high diag-
nostic value. Variations in breast shape and in tissue
morphology within the breast were observed between a
supine and a diagnostic prone MRI of a patient.
Conclusion: The presented supine breast MRI achieved
an image quality comparable to diagnostic breast MRI.
Since supine positioning is common in many clinical
applications such as ultrasound-guided breast biopsy or
breast-conserving surgery, the registration of the supine
images will aid these applications.
Key Words: breast MRI; breast cancer; supine position-
ing; image aid
J. Magn. Reson. Imaging 2011;34:1212–1217.
V
C 2011 Wiley Periodicals, Inc.
1
Sunnybrook Health Sciences Centre, Imaging Research, Toronto, ON,
Canada.
2
Sunnybrook Health Sciences Centre, Department of Surgery, Toronto,
ON, Canada.
3
Sunnybrook Health Sciences Centre, Department of Medical Imaging,
Toronto, ON, Canada.
4
Sentinelle Medical Inc., Toronto, ON, Canada.
Contract grant sponsor: Canadian Breast Cancer Research Alliance
(CBCRA)/Alliance Canadienne pour la recherché sur le cancer du
sein (ACRCS); Contract grant number: 17772.
*Address reprint requests to: P.S., Sunnybrook Health Sciences
Centre, Division of Imaging Research – S 605, 2075 Bayview Ave., To-
ronto, Ontario M4N 3M5, Canada. E-mail: peter.siegler@hologic.com
Received May 23, 2010; Accepted March 9, 2011.
DOI 10.1002/jmri.22605
View this article online at wileyonlinelibrary.com.
V
C 2011 Wiley Periodicals, Inc. 1212
DYNAMIC CONTRAST-ENHANCED (DCE) breast mag-
netic resonance imaging (MRI) exhibits high sensitivity
with somewhat variable specificity (1). It provides an
accurate assessment of the extent of disease and mul-
tifocal disease (2,3).
Breast MRI is commonly performed in the prone
position with the breasts pendant into specialized coil
arrays (4,5). This prone positioning minimizes breast
motion due to respiration and reduces the potential
interference with the beating heart. In addition, the
coil coupling is improved.
While prone breast MRI is well suited for diagnosis,
it is less suitable for use in combination with clinical
applications such as surgery and ultrasound guided
biopsy where the patient is in supine position (6,7).
Given the scale of tissue deformations that are
expected from a prone to supine geometry, achieving
accurate deformation to guide a breast intervention is
computationally challenging, time-consuming, and
subject to substantial errors.
An alternative and more direct approach is the
imaging of the patient in supine positioning. Alderlies-
ten et al (8) showed that this approach substantially
alleviates the need for computationally complex regis-
tration methods while minimizing the potential for
registration errors. However, the body-coil was used
to acquire the supine images, which caused only sub-
optimal image quality.
In contrast to prone breast MRI where the breast is
essentially motionless, respiration causes breast
motion in the supine position (9). Unless efforts are
made for motion compensation, chest wall and breast
motion can degrade image quality.
Breath-holding is impractical for high spatial reso-
lution scans and dynamic breast MRI. Gating meth-
ods (10) can compensate for respiratory motion but
are not time-efficient. Alternatively, phase-encode
reordering methods (11,12) are time-efficient, but
larger motions can degrade the image quality. Techni-
ques that combine reordering and gating were there-
fore introduced (13,14), including the zonal motion-
adapted acquisition and reordering technique
(ZMART) (15) for 3D imaging.
In this work, an MRI apparatus for unilateral breast
MRI in the supine positioning was developed. For
motion compensation, a 3D gradient echo sequence
was modified in accordance with ZMART.
Supine Breast MRI 1213

Figure 1. a: Flexible supine breast coil with four elements used for unilateral supine breast MRI. The black fixture is bend-
able to adapt to the breast geometry. b: Fixture and coil during the positioning of the patient. c: During the examination the
fixture holds the coil above the breast, when the patient is lying supine on the standard patient bed of the MR scanner. In
order to achieve an optimal coil coupling to the desired breast tissue, the flexible imaging coil can be bent as shown in (d)
and (e) or tilted as shown in (f).

MATERIALS AND METHODS Respiratory Motion Compensation

Imaging Coil and Support Structure
All experiments were performed on a whole-body 1.5T
MR scanner (GE Signa Excite, Milwaukee, WI). To
achieve good signal-to-noise ratio (SNR), a unilateral
four-element receive coil (size: 26
23 cm2) for
supine breast MR imaging was built on a flexible Tef-
lon sheet (Fig. 1a) using 6.35-mm wide copper tape.
Coupling between coil loops was minimized by over-
lapping neighboring loops, decoupling capacitors, and
low input-impedance preamplifiers.
A coil fixture was designed to support the coil
above the breast of the patient on the standard bed
of the scanner (Fig. 1b,c). It can be lifted to allow
convenient patient positioning on the patient bed
(Fig. 1b). Once the patient is in position, the fixture
is lowered over the patient’s breast and the coil array
can be positioned above the breast via translation in
all three directions (Fig. 1c). A gimbaled joint (Loc-
Line system, Modular Hose, Lisle, IL) connects the
coil to the fixture and allows arbitrary rotation and
tilting of the coil above the patient’s breast (Fig. 1d–
f). An additional set of gimbaled joints permits the
modification of the coil array shape so that it can
conform as closely as possible to different breast geo-
metries. The coil has no contact with the breast. In
addition, an air gap with a minimal width of 1 cm is
kept between the coil and the breast skin (confirmed
visually). The coil adjustment ensures optimal cou-
pling to the breast tissue, while allowing the breast
to remain in its native supine configuration that is
expected in subsequent surgical or interventional
procedures.
During the entire procedure, a pneumatic respiratory
belt was used to monitor the respiratory state of the
free-breathing patient. The breast images were acquired
with a fast 3D gradient echo sequence (SPGR), which is
frequently used for DCE breast MRI (16). To compen-
sate for the respiratory motion of the breast in the
supine position, the fast 3D SPGR sequence was modi-
fied in accordance with ZMART (15), which is a combi-
nation of phase-encode reordering and gating.
For phase-encode reordering, the actual respiration
state determines which position in the k-space plane
of the two phase-encoding directions (ky and kz) is
acquired. If the respiration state is closer to expira-
tion, a position next to the k-space center is acquired.
As the displacement due to respiration increases, the
acquired position is placed towards the periphery of
k-space. For gating, the acquired data are rejected if
the displacement detected with the respiratory belt
lies above the chosen gating window (spin-conditioned
gating (10)). Phase-encode reordering has no effect on
the scan time. Using gating, the scan time is pro-
longed depending on the gating threshold.
Human Scans
All human supine breast MRI scans were conducted
under a protocol approved by the Human Ethics
Review Board. Images were acquired unilaterally
using the previously described supine breast setup
and a fast 3D SPGR sequence with compensation for
respiratory motion in accordance with ZMART. A
1214
pneumatic respiratory belt was used to monitor the
respiratory state of the free-breathing subject.
When using the supine position, the breast is usu-
ally flattened against the chest wall by gravity, so that
an oblique coronal slice orientation minimizes the
number of necessary slice acquisitions and measure-
ment time. The frequency encoding direction was
always chosen in the left–right direction in order to
avoid artifacts aliasing from the arm and the contra-
lateral breast.
SNR Assessment of Supine Coil Array
To assess the supine coil array, its SNR performance
was compared with that of a routinely used prone coil
array in a volunteer. A set of MRI data was acquired
as order to reconstruct 3D datasets of absolute coil
array SNR using (17,18):
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
SNR ¼ 2pH p= bH Cb; ½1
where p are the scaled complex image data, b the
complex coil sensitivities, and C the receiver noise co-
variance matrix.
After informed consent, the left breast of a human
volunteer was scanned in prone and supine configu-
rations. For each configuration, a set of MRI data was
acquired with a fast 3D SPGR sequence including
image data (flip angle ¼ 30 , field of view [FOV] ¼ 190

190 mm2, Dz ¼ 2 mm, TE ¼ 4.2 msec, TR ¼ 6.5
msec, matrix: 256
256
46, 1 NEX), coil sensitivity
data (TE ¼ 4.2 msec, TR ¼ 5.3 msec, matrix: 32
32

46, 4 NEX), and noise samples (TE ¼ 4.2 msec, TR
¼ 6.5 msec, matrix: 256
256
46, 1 NEX).
In the first configuration, the volunteer was unilat-
erally scanned in the prone position using a conven-
tional eight element High Density Breast Array coil
(GE Healthcare, UK), which had coils in a fixed geom-
etry within the coil housing. Lateral plates were used
to stabilize the breast tissue during imaging. Because
the scan was unilateral, only four coil elements were
used for imaging, which corresponds to the number of
elements of the supine coil array. All images were
acquired in a sagittal slice orientation with frequency
encoding from anterior to posterior to avoid effects of
heart motion and to minimize scan time.
In the second configuration, the volunteer was in
the supine position using the supine imaging appara-
tus (oblique coronal slice orientation, frequency
encoding from left to right). Breast motion caused by
respiration was compensated with ZMART (using
phase-encode reordering and 60% gating window)
based on the signal tracked by the respiratory belt.
Volunteer Study
After obtaining informed consent, unilateral breast
scans were acquired from a free-breathing volunteer
using the supine breast apparatus and a fast 3D
spoiled gradient echo sequence (oblique coronal slice
orientation, frequency encoding from left to right, TE
¼ 4.2 msec, TR ¼ 6.6 msec, flip angle ¼ 30 , FOV ¼
180
180 mm2, Dz ¼ 2 mm and matrix: 256
256

Siegler et al.
50). Datasets were acquired without and with motion
compensation using reordering only.
Patient Study
A supine breast MRI examination was performed on
the affected breast of a patient with previously diag-
nosed breast cancer after obtaining informed consent.
The imaging parameters for the fast 3D SPGR were:
oblique coronal slice orientation, frequency encoding
from left to right, TE ¼ 4.2 msec, TR ¼ 6.5 msec, flip
angle ¼ 30 , FOV ¼ 200
200 mm2, Dz ¼ 2 mm and
matrix: 256
256
46. The first scan showed
adequate motion compensation using only k-space
reordering. As a consequence, no gating was neces-
sary during the examination. After the injection of a
contrast agent (CA; Gd-DTPA), the first scan was
repeated four times. Afterwards one scan was per-
formed with fat suppression achieved using chemi-
cally selective radiofrequency (RF) excitation and
spoiler gradients (19). This fat-suppressed scan
required a longer repetition time of TR ¼ 16.2 msec.
All datasets were viewed in 3D using a software pack-
age (Aegis, Sentinelle Medical, Toronto, Canada),
which allows quick display of 3D datasets from differ-
ent points of view.
The fat-saturated supine images were compared to
diagnostic MR images (bilateral scan, fat saturated,
accelerated by parallel imaging with acceleration fac-
tor R ¼ 2, matrix size ¼ 288
224
84, FOV ¼ 190

190
252 mm3, TE ¼ 3.6 msec, TR ¼ 7.2 msec).
Since the diagnostic scan was performed in a prone
position with sagittal slice orientation, the data were
reformatted to obtain images with a coronal slice ori-
entation for comparison with the supine images.
To account for breast motion between the acquisi-
tions of the supine datasets, all supine 3D datasets
were registered to match the first precontrast dataset
(20). The contrast enhancement was then depicted by
subtraction of the precontrast 3D dataset from the
four postcontrast datasets. Regions of interest (ROIs)
were chosen within the tumor and normal fibrogland-
ular tissue. The average signal inside each ROI was
computed for each subtracted dataset and plotted as
function of time after the injection of the CA.
RESULTS
SNR-scaled maps through the breast of the human
volunteer are shown in Fig. 2 for the conventional
imaging coil (Fig. 2a–d) and the supine coil array (Fig.
2e–h). The SNR of the prone coil array was more ho-
mogenous over the entire breast compared to the SNR
performance of the supine coil array. However, for the
supine coil array the SNR at all points within the
breast was at least as high as for the conventional
coil. Lower SNR values were measured in the tissue
beneath the chest wall (right side in Fig. 2h). The
imaging volume of the supine coil array included also
the breast tissue, which was drawn by gravity to the
side of the torso next to the arm (left side in Fig. 2h).
Supine Breast MRI 1215

Figure 2. SNR-scaled maps through the left breast of a human volunteer using two imaging configurations. a–d: Conven-
tional imaging coil, where data were acquired with prone positioning and with a sagittal slice orientation. Shown are (a) a lat-
eral slice, (b,c) two central slices, and (d) medial slice (spacing between the slices: 16 mm). e–h: Supine coil array, where the
subject was in the supine position with the coil array placed above the breast and an oblique coronal slice orientation. Shown
are (e) a slice near the nipple, (f,g) two slices with increasing distance from the nipple, and (h) a slice near the chest wall
(spacing between the slices: 14 mm). With the supine coil array the SNR over the entire breast volume was at least as good
as the one with the conventional imaging coil. However, the supine coil array showed clear spatial variations in the SNR fol-
lowing the coil profile. The SNR color scale is shown on the right side.

Without motion compensation, the respiratory
motion created typical streaking artifacts along the
phase encode direction in the volunteer study (Fig.
3a–d). These presented themselves as displaced fat
signal outside the breast and within the fibroglandu-
lar tissue. Motion compensation achieved with reor-
dering only (Fig. 3e–h) was quite effective at reducing
these artifacts. Even additional gating did not further
improve the motion compensation.
In the patient study, the tumor was clearly visible
in the supine breast MR images that were acquired
with fat saturation (Fig. 4a–d). However, the use of fat
saturation increased the scan time from 88 seconds
to 190 seconds. When comparing the supine (Fig. 4a–
d) and the reformatted diagnostic images (Fig. 4e–j),
large differences in the orientation and the shape of
the breast and the lesion were observed. Using the 3D
image viewer, the supine data were presented as sur-
face volumes and maximum intensity projections
(MIPs) (Fig. 5a).
The supine images showed some positional changes
of the breast between the scans. The displacements
were mostly in the anterior/posterior direction and
could be successfully corrected using 3D image regis-
tration (20). By subtracting the registered supine
dataset the CA dynamics was clearly depicted. Within

Figure 3. Images of the volunteer with an oblique coronal slice orientation near the chest wall, near the nipple, and in an in-
termediate slice (spacing between slices: 16 mm). The intermediate slice is additionally shown with a hard window for better
visualization of motion artifacts. Without compensation for respiratory motion (a–d), artifacts occurred in the direction of the
in-plane phase-encoding ky (arrows). Motion compensation using reordering only (e–h) successfully prevented motion arti-
facts without increasing the required scan time of 85 seconds.
1216 Siegler et al.

Figure 4. Images of a patient taken in the supine position (a–d) and in the prone position (e–j), where the image data were
reformatted from a sagittal to a coronal slice orientation. The distance between two successive slices was 20 mm in both
sets. The orientation and shape of the breast as well as the position of the lesion (arrows) were fundamentally different in the
two datasets.

the tumor, the signal enhancement showed the typical For the diagnostic scan in prone positioning, the
rapid enhancement followed by a washout phase (solid coils were placed on both sides of each breast as close
line in Fig. 5c). In contrast, the normal fibroglandular as possible. The breast was therefore flattened in the
tissue underwent only minor signal enhancement after regions where it was touching the coil (see sides of
the CA injection (dashed line in Fig. 5c). breast in Fig. 4e–h). In addition, the breast was pend-
ant and extended by gravity. The whole breast geome-
try as well as the relative position of the lesion varied
between the supine and the prone setting (Fig. 4).
DISCUSSION
These variations will cause tumor localization errors
The arrangements of the supine and the prone coils are in the operating room if the prone MR images are
fundamentally different and influence the resulting used to aid clinical procedures.
SNR profiles (Fig. 2). The prone coil array achieved a ho-
mogenous SNR profile over the entire breast (Fig. 2a–d).
In contrast, the sensitivity and thus the SNR profile of
the supine coil array dropped with increasing distance
to the array (Fig. 2e–h). In order to cover the entire
breast, supine imaging coils should therefore always
have a larger volume of sensitivity than conventional
coil arrays used with prone positioning.
The coils were designed to closely conform to the
breast anatomy while not distorting the breast from
its native supine configuration. This is essential for
preserving imaging of diagnostic quality and high
SNR.
Further improvements in coil flexibility can be
achieved with a greater number of coils of smaller
size, which may further increase SNR. This will also
accommodate different breast sizes, the breast tail
and axilla, and thus ensure full coverage of the breast
with high coil sensitivity and SNR.
The k-space reordering alone successfully sup- Figure 5. a: Surface volume (top) and maximal projection
(bottom) of the fat saturated dataset allow a good visualization
pressed respiratory motion artifacts in supine breast
of the 3D volume of the supine breast. b,c: Contrast enhance-
MRI (Figs. 3, 4a–c). An addition of gating was not nec-
ment after CA injection. b: Subtraction image indicating ROIs
essary. Since k-space reordering without gating has within the tumor (solid arrow) and normal breast tissue
no effect on the scan time, the temporal resolution of (dashed arrow). c: Signal enhancement (difference between
supine breast MRI using k-space reordering for respi- images after CA injection and the image before injection)
ratory motion compensation is comparable with cur- inside the two ROIs indicated in (b). The solid line shows the
rent diagnostic breast MRI in the prone position. typical rapid enhancement of tumor with following washout.
Supine Breast MRI
Surface volumes and MIPs allow an easy navigation
through the supine breast (Fig. 5a). Since the arms of
the patient are placed at the side of body during the
MR scan and are outstretched during the surgery,
image registration based on surface information will
be required for image-aided breast-conserving sur-
geries (8).
As the breast was not held by any fixation and thus
remained in its native configuration, image registra-
tion of the dynamic scans was necessary to account
for untracked movements of the patient and changes
in the breathing pattern. The signal-enhancement
curves obtained from supine breast MRI (Fig. 5c)
showed the expected progression in the tumor and
the normal tissue. Supine breast MRI might therefore
have adequate sampling of the contrast uptake curve
for tumor diagnosis.
Nervous and anxious patients may take deeper
breaths and show bigger motion amplitudes during
the supine breast MRI. In these cases, gating in addi-
tion to reordering (ie, ZMART) may be necessary to
achieve sufficient motion compensation. The decision
if gating is necessary can be made based on the qual-
ity of one reordered scan acquired before the injection
of the CA. If gating is used, the resulting temporal re-
solution may no longer be adequate for sampling the
dynamic contrast-enhancement curve.
In conclusion, the presented supine breast MRI
achieved diagnostic image quality and even a depic-
tion of the contrast dynamics were achieved with the
presented supine breast MRI. It mimics the breast
configuration of most clinical procedures more closely
than that of prone breast MRI. This will substantially
improve postimaging registration accuracy.
ACKNOWLEDGMENTS
We thank Michael Pozzobon for assistance with man-
ufacturing of the fixation device, Helen Marshall for
help with the SNR maps, Anne Martel for the 3D
image registration, and Cameron Piron for support
during the coil development.
REFERENCES
1. Peters NH, Borel Rinkes IH, Zuithoff NP, Mali WP, Peeters PH.
Meta-analysis of MR imaging in the diagnosis of breast lesions.
Radiology 2008;246:116–124.
2. Goto M, Ito H, Akazawa K, et al. Diagnosis of breast tumors
by contrast-enhanced MR imaging: comparison between the
1217
diagnostic performance of dynamic enhancement patterns and
morphologic features. J Magn Reson Imaging 2007;25:104–112.
3. DeMartini W, Lehman C, Partridge S. Breast MRI for cancer
detection and characterization: a review of evidence-based clini-
cal applications. Acad Radiol 2008;15:408–416.
4. Konyer NB, Ramsay EA, Bronskill MJ, Plewes DB. Comparison of
MR imaging breast coils. Radiology 2002;222:830–834.
5. Piron CA, Causer P, Jong R, Shumak R, Plewes DB. A hybrid
breast biopsy system combining ultrasound and MRI. IEEE Trans
Med Imaging 2003;22:1100–1110.
6. Carter TJ, Tanner C, Crum WR, Beechey-Newman N, Hawkes DJ.
A framework for image-guided surgery. In: Yang G-Z, et al., eds.
MIAR 2006. LNCS, vol. 4091. Heidelberg: Springer; 2006. p 203–
210.
7. Carter T, Tanner C, Beechey-Newman N, Barratt D, Hawkes D.
MR navigated breast surgery: method and initial clinical experi-
ence. In: Metaxa D, et al., eds. MICCAI 2008, Part II. LNCS, vol.
5242. Heidelberg: Springer; 2008. p 356–363.
8. Alderliesen T, Loo C, Paape A, Muller S. On the feasibility of MRI-
guided navigation to demarcate breast cancer for breast-conserv-
ing surgery. Med Phys 2010;37:2617–2626.
9. Sato Y, Nakamoto M, Tamaki Y, et al. Image guidance of breast
cancer surgery using 3-D ultrasound images and augmented
reality visualization. IEEE Trans Med Imaging 1998;17:
681–693.
10. Ehman RL, McNamara MT, Pallack M, Hricak H, Higgins CB.
Magnetic resonance imaging with respiratory gating: techniques
and advantages. AJR Am J Roentgenol 1984;143:1175–1182.
11. Bailes DR, Gilderdale DJ, Bydder GM, Collins AG, Firmin DN.
Respiratory ordered phase encoding (ROPE): a method for reduc-
ing respiratory motion artefacts in MR Imaging. J Comp Assist
Tomogr 1985;9:835–838.
12. Jhooti P, Wiesmann F, Taylor AM, et al. Hybrid ordered phase
encoding (HOPE): an improved approach for respiratory artifact
reduction. J Magn Reson Imaging 1998;8:968–980.
13. Weiger M, Börnert P, Proksa R, Schäffter T, Haase A. Motion-
adapted gating based on k-space weighting for reduction of respi-
ratory motion artifacts. Magn Reson Med 1997;38:322–333.
14. Sinkus R, Börnert P. Motion pattern adapted real-time respira-
tory gating. Magn Reson Med 1999;41:148–155.
15. Huber ME, Hengesbach D, Botnar RM, et al. Motion artifact
reduction and vessel enhancement for free-breathing navigator-
gated coronary MRA using 3D k-space reordering. Magn Reson
Med 2001;45:645–652.
16. Cron GO, Kelcz F, Santyr GE. Improvement in breast lesion char-
acterization with dynamic contrast-enhances MRI using pharma-
cokinetic modeling and bookend T1 measurements. Magn Reson
Med 2004;51:1066–1070.
17. Kellman P, McVeigh ER. Image reconstruction in SNR units: a
general method for SNR measurement. Magn Reson Med 2005;
54:1439–1447.
18. Marshall H, Devine PM, Shanmugaratnam N, et al. Evaluation of
multicoil breast arrays for parallel imaging. J Magn Reson Imag-
ing 2010;31:328–338.
19. Bottomley PA, Forster TH, Leue WM. In vivo nuclear magnetic
resonance chemical shift imaging by selective irradiation. Proc
Natl Acad Sci U S A 1984;81:6856–6860.
20. Martel AL, Froh MS, Brock KK, Plewes DB, Barber DC. Evaluat-
ing an optical-flow-based registration algorithm for contrast-
enhanced magnetic resonance imaging of the breast. Phys Med
Biol 2007;52:3803–3816.