I.
1
Unenhanced MR Angiography
Martin Backens and Bernd Schmitz
Introduction
In conventional x-ray digital subtraction an-
giography (DSA), administration of contrast agent
is necessary in order to depict blood vessels. After
arterial catheterization and injection of an iodi-
nated contrast agent, two-dimensional projection
images of the lumen of the vessel are acquired
from chosen angles. For every new projection this
procedure has to be repeated. The availability of
3D rotational angiography and CT angiography
may help to overcome this problem but at the ex-
pense of high radiation doses.
Unenhanced MR angiography (MRA) differs
from DSA and other angiographic techniques in
that blood vessels are depicted non-invasively in
the absence of contrast agent injection. Unen-
hanced MR techniques allow the acquisition of 3D
datasets or stacks of 2D images that contain all
vessels in the volume of interest. The acquired im-
ages included in the 3D data set are called “source
images”. Projectional angiographic displays of the
vessel are subsequently reconstructed from the da-
ta using the maximum intensity projection (MIP)
postprocessing algorithm, which generates an-
giogram-like images from the entire dataset or a
subset from any desired viewing angle without the
need for further measurement. Another advantage
of MRA versus x-ray angiography derives from the
fact that extravascular tissue is depicted together
with the vessels, thereby permitting the correlation
of blood flow abnormalities with associated soft
tissue pathologies (Table 1).
Contrast in MR images depends principally on
static tissue parameters: longitudinal relaxation
time T1, transverse relaxation time T2, and proton
density. In addition, the MR signal is sensitive to
flow and movement which frequently leads to arti-
facts in MR imaging. MR angiographic sequences,
however, use flow-induced signal variations to de-
pict blood vessels or even to obtain quantitative
information about blood flow in terms of velocity
and direction.
Unenhanced MRA comprises those MR tech-
niques that rely solely on flow effects. Unlike con-
trast-enhanced MRA (CE MRA) and x-ray angiog-
raphy, which depict the vessel lumen filled with
contrast agent, it is just the movement of blood
that is seen in the unenhanced MR-angiogram.
Flow effects can be grouped into two funda-
mentally different categories:
• Amplitude effects (time-of-flight) Blood flow-
ing into or out of a chosen slice has a different
longitudinal magnetization compared to sta-
tionary spins, depending on the duration of
stay (time-of-flight) in the slice.
• Phase effects Blood flowing along the direction
of a magnetic field gradient is subject to
changes of its transverse magnetization com-
pared to stationary spins.
In principle, both flow phenomena are effective
simultaneously leading either to a decrease or an
increase of the MR signal depending on the type of
sequence used. Appropriate sequence techniques
have been developed which emphasize one of the
flow effects and suppress the other. Typically, MR
angiography techniques are designed in such a
way that flowing blood produces hyperintense sig-
nal while the background signal from stationary
tissue remains largely suppressed (“bright-blood”
angiography). Alternatively, flowing spins can be
Table 1. Advantages of MR angiography
Advantages of MR angiography
No ionizing radiation
Non-invasive
Any projection can be reconstructed from 3D datasets
Depiction of extravascular tissue
Flow quantification in terms of velocity and direction
4 Magnetic Resonance Angiography
Table 2. Techniques of unenhanced MR angiography
Time-of-Flight (TOF) angiography:
Flow changes longitudinal magnetization
Phase-contrast (PC) angiography:
Flow changes transverse magnetization
made to appear hypointense compared to the sta-
tionary background (“black-blood” angiography).
In the clinical setting “bright-blood” MRA is the
more widely accepted of the two techniques. There
are two approaches to performing “bright-blood”
MRA: time-of-flight (TOF) and phase-contrast
(PC) angiography (Table 2).
Although intravenous contrast administration
is not required in TOF and PC MRA, it can be ap-
plied in certain situations to improve vessel con-
trast.
Since unenhanced MRA is based on complex
flow phenomena, physiological conditions of flow
in the vascular territory of interest are of major
importance for the applicability of the method.
Advantageous conditions are found especially in
the vessels of the brain because of the nearly lami-
nar flow in this territory. Moreover, flow is largely
constant during the heart cycle (Fig. 1), making
ECG triggering unnecessary for imaging of brain
vessels. The high velocity of arterial flow (50–100
cm/sec) provides good vessel-background contrast
and moderate acquisition times. In clinical rou-
tine, unenhanced MRA has proven to be a robust
and versatile method for non-invasively imaging
of brain vessels (circle of wilis, sagittal sinus). In
addition, this technique is also suitable for depict-
ing extracranial carotid arteries and short seg-
ments of peripheral vessels (e.g. lower leg).
A major limitation of unenhanced MRA, how-
ever, is a susceptibility to signal loss in areas of tur-
bulent or very slow flow. In severe cases, this may
lead to a misdiagnosis of the pathologic condition
(stenosis, aneurysm). Additionally, TOF and PC
angiography are highly sensitive to motion arti-
facts. Although motion artifacts often arise due to
patient movement because of the need for relative-
ly long acquisition times, they may also occur in
areas of very pulsatile flow such as that occurring
in the carotids, aorta, and especially, in the periph-
eral arteries, and in the thoracic and abdominal
regions due to breathing and heart actions. Where-
as ECG triggering may reduce or eliminate those
artifacts associated with pulsatile flow, the avail-
ability of contrast-enhanced MR angiography (CE
MRA) has largely made unenhanced MRA redun-
dant for most vascular territories outside of the
brain. However if unenhanced MRA techniques
are performed, a thorough understanding of the
underlying physical and technical mechanisms is
prerequisite to performing imaging and to cor-
rectly interpreting the acquired angiograms.
Understanding Flow Effects
Outflow-related Signal Loss
(washout effect, T2 flow void)
When images are obtained with a spin-echo (SE)
pulse sequence, blood flowing at a high velocity
perpendicular to the imaging plane produces a
weaker signal than the surrounding stationary tis-
sue. This phenomenon is caused by the washout of
flowing spins from the slice during the imaging
process.
Spin-echo techniques are characterized by a se-
quence of slice-selective 90° and 180° radio fre-
quency (RF) pulses. Only those tissue components
that are affected by both pulses can provide an MR
signal. Moving material, such as blood in the ves-
sels, flowing through the excited slice at a suffi-
Fig. 1. Comparison of flow pro-
files in the external carotid artery
(ECA), the internal carotid artery
(ICA) and the middle cerebral ar-
tery (MCA). There is very pulsatile
flow in the ECA. However, in the
intracranial vessels, the variation
of flow during heart cycle is much
less pronounced [Courtesy of
Steffi Behnke, MD, Dept. of Neu-
rology, Saarland University Hospi-
tals]

ciently high velocity, is affected by only one of
these pulses, and therefore does not contribute
to the MR signal. This is the so-called flow void
(Fig. 2).
The intensity of the vascular signal declines
with
• decreasing slice thickness, s,
• increasing echo time, TE,
• increasing flow velocity, v.
If the blood flow velocity is so high that all
spins leave the slice between the 90°- and 180°-
pulses (v ≥ s/(TE/2)), there will be no signal at all
and the vessel will appear dark.
Spins flowing within the imaging plane are not
affected by this phenomenon.
I.1 • Unenhanced MRA 5
Fig. 2. In spin-echo sequences, blood
flowing out of the measured slice in
the time between 90°- and 180°-ra-
diofrequency pulses leads to signal
loss
The washout effect is observed only for SE se-
quences and is most pronounced on T2- weighted
imaging because of the long echo times used. With
gradient-echo (GRE) techniques, the echo is refo-
cused without a 180° pulse simply by reversing the
imaging gradients. Since only one RF pulse is
needed to form an echo, the washout effect does
not occur.
With standard SE sequences the washout effect
provides valuable and reliable information about
blood flow. The absence of a flow void on T2-
weighted imaging should be considered as indica-
tive of very slow flow or even occlusion of the ves-
sel (Fig. 3). On the other hand, occlusion of the ves-
sel can be excluded if a flow void is present.
Fig. 3. Axial T2-weighted
spin-echo image of the
pons region. Left: Missing
flow void in the basilar ar-
tery indicates thrombosis
of the vessel. Right: After
thrombolysis, the vessel
(arrow) appears dark due
to restored flow
6 Magnetic Resonance Angiography
Inflow-related Signal Enhancement
(Inflow Effect)
Although the signal of blood flowing quickly out
of the measured slice is reduced with SE se-
quences, under certain circumstances the opposite
effect may occur: spins flowing into the slice may
generate a higher signal than the surrounding tis-
sue. This effect is referred to as inflow enhance-
ment.
On T1-weighted imaging, contrast is generated
by repeated RF pulses that are applied with a time
interval (repetition time, TR) that is shorter than
the T1 relaxation time of the tissue (typically TR <
700 msec). As a result, the tissue components are
saturated unequally, depending on their individual
T1 times. This is the basis of T1-weighted image
contrast (Fig. 4). Irrespective of flow effects, blood
in the vessels would appear hypointense on a nor-
mal T1-weighted image due to its relatively long
T1 time.
The signal emitted by the tissue diminishes
when the TR is reduced. With GRE sequences, rep-
etition times shorter than 50 msec can be
achieved. This allows the majority of non-moving
spins to become saturated, thus minimizing the
background signal.
Spins outside the excited slice (or volume) are
not influenced by the RF pulses. Consequently,
blood entering into the slice being imaged is fully
relaxed, experiencing not more than a few excita-
tions on its way through the slice. As a result, flow-
ing blood gives rise to considerably higher signal
intensity relative to that of the saturated spins in the
stationary tissue. This effect is called “inflow en-
hancement” or “flow-related enhancement” (Fig. 5).
The signal intensity of flowing blood increases
with:
• decreasing slice thickness s,
• increasing flow velocity v.
If the blood flow velocity is so high that all ves-
sel spins are replaced by unsaturated spins in the
time interval TR (i.e. v > s/TR), flow enhancement
is maximal and the vessel appears bright on a gray
or black background (Fig. 6).
Although the inflow effect occurs both with SE
and GRE sequences, SE sequences are not practical
for the TOF method because the competing
washout effect (see above) tends to overbalance
the inflow effect at higher flow velocities, leading
to decreased flow signal.
Consequently, flow-related enhancement using
GRE sequences to produce bright-blood images is
the basis of time-of-flight angiography.
Fig. 4. With a very long repeti-
tion time TR, the magnetization
fully relaxes yielding maximum
signal strength M0. Shortening of
the TR leads to partially saturated
magnetization and therefore de-
creased signal

Fig. 6. Brain vessels, axial slice. Due to the inflow effect,
there is high contrast between the vessels and the sur-
rounding tissue
Phase Effects
Phase effects concern the transverse component of
the magnetization. They occur whenever spins are
moving in the presence of magnetic field gradi-
ents, as are applied for spatial encoding of the MR
signal.
Magnetic field gradients provoke a change in
the Larmor frequency depending on gradient
strength and spin position. A gradient pulse of cer-
tain length and amplitude therefore induces a
phase shift of the transverse magnetization, which
can be compensated by a second gradient pulse
I.1 • Unenhanced MRA 7
Fig. 5. Inflow effect: In the interval between two RF
excitations, the blood in the vessel is replaced by
“fresh”, unsaturated blood, while the stationary tis-
sue in the slice is saturated due to the short TR
Fig. 7. Spins moving along bipolar magnetic field gradients experience a
phase shift Φ of their transverse magnetization proportional to the gradient
A, the time interval t between the pulses, and the velocity v of the spins along
the gradient direction. In contrast, the phase shift for stationary spins is zero
with identical strength and duration but opposite
sign. Thus, for stationary spins the net phase shift
is zero. In contrast, the same gradients applied on a
flowing spin generate a non-zero phase shift. Since
the spins change their position during the bipolar
gradient application, the second gradient pulse is
no longer able to completely compensate for the
phase shifts induced by the first gradient. The re-
maining phase shift Φ is proportional to the veloc-
ity component v of the spins along the gradient di-
rection (Fig. 7).
On standard MR imaging, this flow-induced
phase shift causes a spatial misencoding of the sig-
8 Magnetic Resonance Angiography

Optimal reduction of flow-induced phase ef-

fects can be achieved by combining GMR with as
short a TE as possible, in order to reduce the time
available for spin dephasing. Short echo times also
diminish the impact of pulsatile blood flow and
turbulence.

Fig. 8. Blood vessel exhibiting laminar flow with a parabolic flow Time-of-Flight (TOF) Angiography
profile. The flow velocity is indicated by the length of the arrows
showing an increase from the border towards the center Techniques
The contrast mechanism of TOF MRA is based on
the inflow effect. Fully relaxed blood entering the
nal leading to ghost artifacts that are typically
measured volume behaves as an endogenous con-
found in the phase-encoding direction.
trast agent, by producing a bright signal. The bright
Spins in a blood vessel are moving with differ-
depiction of flowing blood, however, requires the
ent velocities. Often, a parabolic flow profile is
use of flow rephasing techniques (GMR) in order to
found (Fig. 8). Spins moving faster experience a
overcome the effects of spin dephasing due to
larger phase shift than those moving more slowly. If
transverse magnetization. TOF MRA using GRE se-
there is a velocity distribution inside a voxel, phase
quences has several advantages: Firstly, GRE se-
dispersion (intra-voxel dephasing) occurs resulting
quences are not affected by the wash-out phenome-
in decreased signal in the blood vessel (Fig. 9). The
non that diminishes the signal of fast flowing blood
extent of spin-dephasing depends on the strength
when using SE techniques. Secondly, GRE tech-
and time interval of the gradient pulses, as well as
niques permit the use of short repetition times (TR
the distribution of spin velocities. When complex
< 40 msec), which are needed to efficiently saturate
flow patterns are encountered, for example in ves-
stationary tissue. Thirdly, echo times can be kept
sels with turbulent flow, there may be a very broad
short (TE < 5 msec), thus further reducing spin de-
spectrum of velocities within a voxel, leading to to-
phasing. Generally, it is advisable to apply in-phase
tal signal loss in the vessel (Fig. 9 c).
echo times in order to avoid opposed-phase effects
Using additional gradient pulses of appropri-
at the vessel walls that would impair the depiction
ate amplitude and duration, flow-induced phase
of small vessels. Finally, GRE techniques are charac-
shifts can be compensated, thus eliminating any
terized by short acquisition times which are impor-
signal loss. This technique is called “gradient mo-
tant when acquiring volume (3D) datasets.
tion rephasing (GMR)” or just “flow compensa-
TOF techniques can be divided into three
tion” [1]. However, GMR is normally restricted to
groups (Fig. 10):
first-order movements, i.e. spins that move at a
constant velocity. Turbulent flow and effects of ac- • sequential 2D multi-slice method,
celeration cannot be completely compensated by • 3D single-slab method,
GMR. • 3D multi-slab method.

Fig. 9a-c. a If all spins in a

voxel have the same veloci-
ty the signal is maximal. b If
there is velocity distribu-
tion, loss of phase coher-
ence occurs resulting in de-
creased signal intensity. c If
a b c phase dispersion is total, the
signal intensity is zero

With 2D techniques, the vessel is imaged by se-
quentially scanning multiple thin slices. This
method has two advantages in comparison to the
interleaved multi-slice technique: Firstly, very short
TR times can be used which boost the inflow effect,
and secondly, partially saturated blood is hindered
from flowing from one slice to another. The
method guarantees sufficient inflow enhancement
even in vessels with a very slow flow and produces
constant vessel-background contrast in the cov-
ered region, because each slice is an entry slice [2].
Problems arise with 2D TOF MRA if the vessels
to be imaged do not flow in a perpendicular direc-
tion to the imaging plane. If the vessels run partly
inside the slice (i.e. in-plane) or return to the slice
in a looped form, then signal loss may occur due to
the partial saturation of flow.
In order to achieve a sufficient signal-to-noise
ratio, a slice thickness of at least 2-3 mm is neces-
sary. However, this results in reduced spatial reso-
lution and increased spin dephasing due to the
larger voxel size required. 2D slices do not have
rectangular RF profiles and therefore exhibit sig-
nal variations at the edges that can lead to step-
like artifacts in maximum intensity projection
(MIP) reconstructions. However, this effect can be
largely overcome by overlapping the slices.
In vessels exhibiting very pulsatile flow, the ex-
tent of inflow enhancement varies during the heart
cycle. The periodic change of inflow enhancement
generates ghost images of the vessels. Saturation
of blood spins can occur due to slow flow in the di-
astolic phase. The use of ECG triggered 2D TOF se-
quences may overcome many of these problems by
confining data acquisition to the phase of maxi-
mum inflow. Thus, blood signal is enhanced and
ghost artifacts are eliminated. By synchronizing
data acquisition with the heart cycle, the vessel is
mapped in each slice with equal intensity. Unfortu-
nately, a drawback of this approach is the pro-
longed acquisition time.
I.1 • Unenhanced MRA 9
Fig. 10. TOF-techniques:
2D multi-slice, 3D single-
slab, 3D multi-slab
In 3D TOF MRA, the entire imaging volume,
usually 30 – 60 mm thick, is excited simultaneously
and then partitioned into thin slices by an addi-
tional phase encoding gradient along the slice-se-
lect direction [3]. 3D TOF MRA has the advantage
of high spatial resolution together with high sig-
nal-to-noise ratio, thereby facilitating the improved
depiction of particularly small vessel structures.
The technique allows slices of less than 1 mm
thickness and isotropic voxels to be acquired easily.
One major problem of the 3D technique, how-
ever, is the progressive saturation that occurs when
blood flowing through the volume is subjected to
repeated RF pulses. As a result, the signal intensity
decreases continuously in the direction of flow.
The extent of saturation depends on the length of
time in which the blood stays inside the volume. In
slow flow vessels, signal begins to diminish when
only a short distance has been covered. Conversely,
in faster flowing blood the signal remains visible
for a greater distance. Consequently, the maximum
volume thickness should be kept as small as possi-
ble, matched to the size of the vessel region of in-
terest. A reduction of saturation can also be
achieved by increasing the TR (see Fig. 5).
Larger vessel sections can be investigated by
subdividing the volume of interest into several
thin 3D slabs that are acquired sequentially [4].
One such multi-slab technique, which retains the
advantages of 3D TOF yet has reduced saturation
effects like 2D TOF, is called multiple overlapping
thin slab acquisition (MOTSA). Generally, the cho-
sen slab thickness has to be small enough to avoid
saturation within the slabs. However, adjacent
slabs must overlap by about 20 – 30% in order to
compensate for signal attenuations arising at the
slab edges due to non-rectangular excitation pro-
files. This results in compromised time efficiency
and longer overall acquisition times.
The advantages and disadvantages of 2D and
3D TOF MRA are summarized in Table 3.
10 Magnetic Resonance Angiography
Table 3. Comparison of 2D TOF and 3D TOF angiography
2D TOF
Strong inflow effect, minimal saturation
• sensitive even to slow flow (veins)
• sensitive to rather fast flow (arteries)
Relatively poor signal-to-noise ratio
Short scan times
Relatively thick slices
• suitable for large vessels
• suitable for small vessels
Poor in-plane flow sensitivity
• for straight, unidirectional flow
Long echo times
Step artifacts at the vessel wall
Table 4. Options to improve TOF angiography
Orientation of slices or volume perpendicular to flow direction
2D for slow flow, 3D for fast flow
3D multi-slab for larger vessel sections
Spatial presaturation to isolate arteries and veins
Use of minimum TE reduces signal loss due to spin dephasing
TONE pulse reduces saturation effects in 3D TOF
Magnetization transfer (MTC) and fat suppression improve vessel contrast
Frequently, 2D TOF MRA is favored for imag-
ing veins because of the high sensitivity to slow
flow. 3D TOF MRA, on the other hand, is more ap-
propriate for fast arterial flow and for those cases
in which high spatial resolution is required.
Optimization (Table 4)
In TOF MRA, arteries and veins are often dis-
played simultaneously, since the inflow enhance-
ment is equally effective for directly opposed flow
directions. As a result, there may be an interfering
overlap of arterial and venous vessels in the MIP
reconstruction which hampers the detection or as-
sessment of vessel lesions. This drawback can be
overcome by applying additional presaturation
slabs that saturate and dephase spins before they
enter the image slice. A selective arteriograph is
generated if the presaturated area is placed distal
to the imaging volume, i.e. at the entry side of the
veins. In this case, inflowing venous spins are satu-
rated in a manner similar to stationary spins and
therefore do not produce a bright signal. Con-
versely, presaturation of arterial inflow permits the
selective depiction of veins. In order to achieve
sufficient suppression even of fast flowing blood,
presaturation slabs of several centimeters thick-
ness must be positioned close to the imaging vol-
3D TOF
More saturation effects
High signal-to-noise ratio
Poor background suppression
Thin slices, allows isotropic voxels
Better than 2D TOF for tortuous vessels
Short echo times, less dephasing
Smoother vessel walls
ume. In case of 2D TOF sequences traveling presat-
uration slabs may be employed that move along
with the imaged slice but at a constant distance
from the actually imaged slice.
For all TOF techniques, it is important to posi-
tion the imaging volume (slices or slabs) perpendi-
cular to the vessel to minimize the length of the
vessel section in the slice or slab and to reduce the
saturation effects. As intracranial vessels run in dif-
ferent directions, it is not possible to simultaneous-
ly orientate the imaging volume perpendicular to
all vessels. Consequently, the 3D multi-slab tech-
nique has proven to be the best method for investi-
gating brain arteries. In order to guarantee optimal
coverage of the arterial vessel tree with minimal
slab thickness it is advisable to tilt the slabs slight-
ly from the axial to the coronal orientation (Fig.
11). The resulting MIP reconstruction gives excel-
lent visualization of the brain arteries (Fig. 12).
The depiction of venous brain vessels using the
TOF technique can be optimized by choosing a
sagittal slice orientation tilted slightly towards
coronal and axial directions (Fig. 13). As a result of
this spatial orientation, saturation of blood in the
sagittal sinus that could occur due to long-range
flow within one single slice is prevented. Inflow en-
hancement is sufficient for all veins, regardless of
their flow direction (Fig. 14).
The flip angle of a GRE sequence considerably
 I.1 • Unenhanced MRA 11

Fig. 11. Typical position of a 3D TOF multi-slab. The presaturation Fig. 12. MIP reconstruction from a 3D multi-slab TOF MRA dataset
slab above the imaging volume suppresses the signal of venous flow of the brain arteries

Fig. 13. Imaging volume of a 2D TOF MRA of the brain veins. By tilting the sagittal slices in an axial and coronal orientation, saturation of
venous blood in the sinus sagittalis is avoided. The saturation slab located beneath the imaging volume suppresses the arterial flow signal

Fig. 14. MIP reconstruction of a 3D TOF MRA dataset of the venous brain vessels
12 Magnetic Resonance Angiography
influences the vessel-background contrast. Large
flip angles generate a high signal at the entry side
of the blood, but also provoke rapid signal decrease
along the course of the vessel due to saturation.
Small flip angles produce a lower vessel-back-
ground contrast but less saturation. 2D TOF typi-
cally uses flip angles in the range of 30° - 70°,
whereas 3D TOF employs lower angles between 15°
and 20° to reduce saturation. In the technique re-
ferred to as TONE (Tilted, Optimized, None-Satu-
rating Excitation) [5], the flip angle is varied lin-
early in the slice direction, beginning with small
values at the entry side and ending with high val-
ues at the exit side of the volume (Fig. 15). In this
way, saturation across 3D slabs can partially be
compensated to achieve a more uniform signal dis-
tribution along the course of the vessel. The extent
of the flip angle variation depends on flow direc-
tion, flow velocity (slow, medium, fast) and slab
thickness. Therefore, sequence protocols that uti-
lize TONE are optimized for specific vessel regions.
Further suppression of background signal can
be achieved using magnetization transfer contrast
Fig. 15. By linearly varying the flip angle across the
volume (TONE), the progressive saturation of the flow
signal occurring with 3D TOF MRA can be reduced
(MTC) [6, 7]. By applying a radio pulse at a much
different frequency from that of water resonance,
protons of motionally restricted macromolecules
are saturated. In contrast mobile water protons
that generate the MR signal in vessels and tissues
are not affected by this off-resonance pulse. By
cross-relaxation and chemical exchange, the satu-
ration is transferred to the neighboring free pro-
tons, thus significantly reducing the measured MR
signal. This effect can be seen in the gray and white
brain matter (reduction of signal by 15-40%) but
not in blood. Consequently, the contrast between
the vessel and the surrounding tissue is enhanced.
This is particularly useful for depiction of small
vessels and slow flowing blood which are made ap-
preciably more visible.
Unfortunately, fat does not exhibit magnetiza-
tion transfer. Therefore, to avoid enhanced fat con-
trast, MTC should be used in combination with fat
suppression techniques. Effective suppression of
the signal of fat may be achieved either by direct
frequency-selective saturation of the fat signal or
by selective excitation of water (Fig. 16). The dis-
Fig. 16. Left: non-selective
excitation. Right: fat suppres-
sion using selective water ex-
citation
 I.1 • Unenhanced MRA 13

Fig. 17. TOF MRA pre (left)

and post (right) administra-
tion of gadolinium contrast
agent. After contrast injection,
the suppression of veins is no
longer effective

advantage of fat saturation in comparison with
water excitation is that water protons instead of fat
protons could be accidentally affected by the pre-
saturation pulse in regions outside the shim vol-
ume (i.e. in the less homogeneous magnetic field)
resulting in unwanted reduction of inflow effect.
Unlike unenhanced MRA, contrast-enhanced
MRA (CE MRA) is usually achieved through the
application of an exogenous gadolinium-based
MR contrast agent which shortens the T1 relax-
ation time of proton spins in its immediate vicini-
ty. After intravenous administration, the T1 of
blood is strongly reduced, thereby counteracting
problems associated with the saturation of spins
flowing across the volume. As a consequence, the
contrast between blood vessels and surrounding
tissues is improved, primarily for small vessels and
those with slow flow. There are, however, two ma-
jor disadvantages arising from contrast agent ad-
ministration: First, if soft tissue in the imaging
area shows considerable contrast enhancement,
the background signal will also be enhanced. Sec-
ond, due to the fast T1 relaxation of blood protons,
suppression of either the signal from the arteries
or veins using presaturation pulses will malfunc-
tion (Fig. 17). In such cases, evaluation of source
images or multiplanar reconstructions may yield
more information than MIP reconstructions.
Pitfalls (Table 5)
Tissues with very short T1 relaxation times, such
as fat, methemoglobin after intracranial hemor-
rhage or in fresh thrombi, and contrast enhancing
structures, pose a problem for TOF MRA. As a re-
sult of inefficient saturation, these tissues often
produce a very bright signal which may render
them undistinguishable from flowing blood. For
example, methemoglobin in a hemorrhage (Fig.

Table 5. Pitfalls with TOF MRA

Methemoglobin in thrombosed vessels (cavernous: sinus thrombosis) may mimic blood flow (i.e., vessel patency)
Workaround: compare MIP with pre-contrast T1 images or use phase contrast MRA
Short T1 tissues (fat, bleeding, tissue that take up contrast) may simulate vessels
Pulsation artifacts in CSF may simulate vessel lesions
Signal loss occurring with turbulent or very slow flow causes overestimation of stenosis and artifacts in the depic-
tion of aneurysms
Signal loss due to susceptibility artifacts (coils, clips)
Signal loss in case of in-plane flow (2D) or slow flow (3D)
Overlap of arteries and veins after contrast administration, particularly in intracranial MRA
14 Magnetic Resonance Angiography

Fig. 18. Methemoglobin in

an intracranial hemorrhage.
Left: MIP reconstruction.
Right: T1-weighted image

18) may be mistaken for an aneurysm while oc-

cluded vessels, such as sinus thrombosis (Fig. 19),
may be misinterpreted as perfused. Confusions of
this kind can be avoided by comparing the TOF
data with precontrast T1-weighted acquisitions.
As shown in Fig. 20, pulsating cerebrospinal
fluid (CSF) may cause artifacts on TOF MRA. In
this case the 3D TOF multi-slab technique pro-
vokes strong inflow enhancement at the entrance
of the upper slab, generating a bright signal of the
CSF in the cerebral aqueduct of Sylvius on the
sagittal reconstruction of the source images (Fig.
20 c). In the projection reconstructions obtained
by MIP postprocessing (Fig. 20 a, b), this structure
may be misinterpreted as a vessel lesion.
Distal to a stenosis or near arteriosclerotic al-
terations of the vessel wall, turbulent and acceler-
ating flow may lead to signal loss due to spin de-

Fig. 19. Methemoglobin in sinus vein thrombosis. Bright depic-

tion of the occluded veins in the unenhanced T1-weighted image

Fig. 20a-c. Artifact (arrows) in the sagittal (a) and axial (b) MIP
reconstructions, caused by pulsating cerebrospinal fluid (c, arrow)
c in the cerebral aqueduct as shown in a reconstructed sagittal slice

phasing. The degree of stenosis may therefore be
overestimated (Fig. 21). The problem can be re-
duced, if not entirely eliminated, by the use of very
short TE values. In addition, the low flow velocities
that may exist in severely stenotic vessels may lead
to increased saturation effects and hence reduced
flow contrast.
Similarly, the evaluation of aneurysms is sus-
ceptible to artifacts caused by turbulence or very
slow flow.
Further pitfalls associated with TOF MRA de-
rive from the fact that GRE sequences in general
are sensitive to distortions of the magnetic field
originating from metallic implants (clips, coils,
etc.). Likewise, signal loss due to magnetic suscep-
tibility artifacts may simulate an interrupted ves-
sel, while signal decrease resulting from saturation
effects is always a major problem with TOF MRA.
In 2D TOF MRA, signal loss should be consid-
ered if vessels are running within the imaging
plane. It is therefore essential to place the slices
perpendicular to the vessel. In 3D TOF MRA, satu-
ration occurs if thick slabs are applied to vessels
with slow flow.
Phase Contrast Angiography
Whereas phase effects are suppressed as fully as
possible in TOF MRA, it is the flow-induced phase
shift of the transverse magnetization that is em-
ployed to image blood vessels with phase contrast
techniques. There are two ways of using phase ef-
fects to selectively depict blood flow:
• magnitude contrast (rephased-dephased)
method
• phase contrast method
I.1 • Unenhanced MRA 15
Fig. 21. Overestimation of the length of a
stenosis due to turbulence
Today, magnitude contrast is only rarely used.
Conversely, techniques that in a stricter sense are
referred to as “phase contrast”, have gained greater
importance both as an imaging method and as an
accurate approach to measuring blood flow veloc-
ity and direction.
Magnitude Contrast Technique
The concept of magnitude contrast angiography is
analogous to x-ray digital subtraction angiography
(DSA). The basic idea is to acquire two datasets:
one flow-rephased and one flow-dephased [8].
First a flow-compensated measurement is per-
formed using GMR in order to image flowing
blood with high signal intensity. In the second ac-
quisition, flow-sensitizing bipolar gradient pulses
are applied specifically to induce velocity-depend-
ent phase shifts of moving spins. If the flow-sensi-
tizing gradient is strong enough, the spins within a
voxel possessing different velocities may totally
dephase resulting in dark vessel signal (Fig. 22).
Since stationary tissue appears the same in both
acquisitions, subtraction of one dataset from the
other results in the signal of the stationary tissue
being cancelled leaving only the moving blood as
visible (Fig. 23). Interleaving the acquisition of
both datasets can diminish the impact of motion
artifacts on the subtraction process.
The signal intensity in the subtracted image de-
pends only on the velocity component along the
flow-sensitizing gradient which is normally ap-
plied in the frequency-encoding direction. There-
fore, the imaging volume should be oriented in
such a way that the main flow direction in the ves-
sel of interest is parallel to the read-out direction.
16 Magnetic Resonance Angiography
Information about all three orthogonal flow di-
rections can be obtained by repeating the flow-en-
coded acquisition with altered gradient orienta-
tions. As a result, a total of four acquisitions has to
be performed (one rephased, three dephased),
which impacts on the overall acquisition time.
Although not in widespread use, magnitude
contrast MRA can be considered applicable for im-
aging of peripheral vessels (arm, leg) since it al-
lows larger sections of arteries to be visualized. If
flow is unidirectional, a single pair of rephased
and dephased acquisitions is sufficient, thereby re-
ducing the overall acquisition time.
Magnitude contrast MRA requires a spectrum
of flow velocities within a voxel. Laminar flow with
its parabolic flow profile is therefore readily de-
tected. The signal acquired is a direct result of the
velocity distribution in each voxel, ensuring com-
plete background suppression. The method is well
adapted for depicting slow flow with good spatial
resolution, covering larger sections of vessels. One
disadvantage arises from the fact that this tech-
nique does not provide any information on flow
direction or flow velocity.
Fig. 22. Left: Using flow com-
pensation (GMR), the phases of
all spins regardless of their ve-
locity are aligned. Thus, the sig-
nal is maximal, the vessel ap-
pears bright. Right: Using strong
flow-sensitizing gradients, com-
plete dephasing of the spins oc-
curs due to their different veloc-
ities. In this case, the signal is
nulled, the vessel appears black
Fig. 23. Principle of magni-
tude contrast angiography. First
a flow-rephased dataset is ac-
quired followed by a flow-de-
phased acquisition. After pixel
by pixel subtraction (similar to
x-ray digital subtraction angiog-
raphy (DSA)) only the signal of
flowing blood remains, while
background signal is nulled
Phase Contrast Technique
Similar to magnitude contrast, phase contrast an-
giography is based on the acquisition of two
datasets that differ in the phase of moving spins
[9, 10]. At first, a flow-rephased sequence (S1) is ap-
plied which defines the phase of transverse mag-
netization under conditions of total flow compen-
sation. The second measurement (S2) is flow-sensi-
tive, utilizing special flow-encoding gradients to
provoke a measurable phase shift. Contrary to the
magnitude contrast technique, the chosen gradi-
ent is weak enough to avoid complete phase dis-
persion arising from the velocity distribution of
the spins. Complex subtraction of the two datasets
S1 and S2 yields the phase difference Φ as well as
the difference vector ΔS, both of which depend on
the velocity component of the spins along the flow-
encoding gradient (Fig. 24).
There are two different approaches to utilizing
flow-induced phase shifts to generate angiograph-
ic images. In the so-called phase map images, it is
the phase difference Φ that is depicted as signal in-
tensity. The sign of the phase shift encodes the

flow direction. Non-moving tissue appears as a
medium shade of gray while flowing blood is ei-
ther brighter or darker, according to the direction
of flow. As the intensity of each pixel is linearly
proportional to flow velocity, phase images are
particularly well suited for flow quantification and
for identifying flow direction.
In the second approach Magnitude images,
which reveal the length of the difference vector ΔS,
are applied to anatomically image the vessels.
However, while the brightness in each pixel is a
measure of the local flow velocity, there is no in-
formation about the flow direction.
The difference vector ΔS increases with rising
spin velocity, reaching a maximum at Φ=180°. The
corresponding critical velocity is called flow sensi-
tivity or velocity encoding (venc) and is deter-
mined by the strength of the bipolar flow-encod-
ing gradients. The manufacturers of MR scanners
provide a set of sequences adapted to different ve-
locity ranges. In clinical practice, it is important to
estimate the maximum flow velocity expected to
occur in the vessel in advance in order to choose
the phase contrast sequence with an adequate venc
value.
I.1 • Unenhanced MRA 17
Fig. 24. Principle of phase con-
trast angiography: A flow-com-
pensated (S1) and a flow-sensitive
(S2) dataset are acquired. Com-
plex subtraction yields the phase
difference Φ and the difference
vector ΔS. Both quantities de-
pend on the local flow velocity
and can be used for generating
flow images
Fig. 25. In the phase image, the
value of the phase shift Φ is de-
picted as signal intensity. In the
range from –venc to +venc, the
phase shift is directly proportion-
al to the flow velocity. Flow that
is faster leads to a turnover of the
phase (aliasing)
With phase contrast, only those velocities rang-
ing between –venc and +venc, corresponding to
phase shifts between –180° and +180°, can be
uniquely detected. If the flow velocity exceeds the
venc value, there is an abrupt change of signal in-
tensity in the phase image from bright to dark or
vice versa (Fig. 25). Blood flow that provokes a
phase shift of 190° cannot be distinguished from
oppositely directed flow that generates a phase
shift of –170°. This ambiguity is called aliasing.
In the magnitude image, maximum intensity is
reached when the flow velocity equals the venc val-
ue (Fig. 26). At higher velocities, the signal intensi-
ty begins to decrease again. If the flow velocity in a
voxel is exactly twice the venc value, no signal em-
anates from the voxel, and the vessel appears to be
interrupted.
As an example, for a venc value of 40 cm/sec
spins flowing at a rate of 40 cm/sec provoke a phase
shift of 180° yielding maximum signal intensity.
Spins that flow at 60 cm/sec possess a phase shift
higher than 180°. Therefore, on the magnitude im-
age they appear less bright than the spins flowing at
40 cm/sec, but equally as bright as spins flowing at
a rate of 20 cm/sec. Likewise, if there are spins flow-
18 Magnetic Resonance Angiography
ing at a rate of 80 cm/sec, they will provide zero
signal and appear as stationary spins.
The phase contrast method is sensitive only for
the velocity component along the flow-encoding
gradient. In order to obtain information on all flow
directions, one dedicated flow-encoding gradient
for each orthogonal direction of space is required.
Thus, a phase contrast sequence comprises a total of
four acquisitions: one flow-compensated measure-
ment and three flow-encoded acquisitions in the x-,
Fig. 26. In the magnitude image, the
length of the difference vector DS is
depicted as signal intensity. Brightness
is maximal at the venc value and de-
creases at higher flow rates. If the ve-
locity is twice the venc value, the signal
intensity is zero
Fig. 27. Complex subtraction
of the flow-encoded and flow-
compensated datasets yields
phase- and magnitude images
for the three directions in
space. By combining the mag-
nitude images, a sum magni-
tude image can be obtained
that gives a vessel image that
is independent of flow direc-
tion
y-, and z-directions. Interleaving the four datasets
can reduce artifacts caused by patient motion.
Phase and magnitude images of the flow com-
ponents in the three orthogonal directions are ob-
tained by complex subtraction of flow-encoded
and flow-compensated datasets (Fig. 27). The three
magnitude subtraction images can be added to ob-
tain a sum magnitude image that depicts blood
flow with bright signal regardless of flow direction
(Figs. 27, 28).

Optimization (Table 6)
The correct choice of the flow sensitivity of a
phase contrast sequence is critical to the quality of
vessel depiction. If the chosen venc value is too
high, the acquired signal-to-noise ratio is poor due
to the small difference between signals. Converse-
ly, venc values that are two low lead to aliasing ar-
tifacts and flow voids which may be misconstrued
as stenosis. Typical flow velocities in some larger
vessels, which may indicate the appropriate venc
values to use, are given in Table 7.
Blood flow in arteries and veins is rarely con-
stant during the heart cycle. Therefore, given that
phase contrast sequences require several minutes
of acquisition time, it is not the maximum but the
Table 6. Options for improving phase contrast MRA
Adapting flow sensitivity (venc) to maximum flow velocity
Encoding different flow velocities (multivenc) or different flow directions
Contrast agent improves flow signal
2D acquisition provides one single projection within a short acquisition time,
3D acquisition permits MIP postprocessing
ECG triggering can be applied in cases of pulsatile flow
Presaturation pulses can separate arteries and veins
I.1 • Unenhanced MRA 19
Fig. 28. 2D phase contrast angiography: phase im-
ages encoded for the x, y, and z-directions and the
corresponding sum magnitude image
mean flow velocity averaged over the heart cycle
that is relevant. In the brain the vessels are highly
tortuous and thus blood flow is hardly ever direct-
ed along one single phase encoding direction.
Hence, in practice, the maximum signal intensity is
obtained when the chosen venc value is about half
the maximum flow velocity, as demonstrated in
Fig. 29.
Due to the irregular course and position of ves-
sels in the brain, it is reasonable to encode one sin-
gle flow sensitivity in all three orthogonal direc-
tions. Conversely, in the peripheral arteries, there
is one main flow direction, but a great variation of
flow velocities. Consequently, in the peripheral ar-
teries a “multi-venc” measurement can be per-
formed in which three flow velocities (e.g. 30, 60,
20 Magnetic Resonance Angiography

Table 7. Flow velocities in some large vessels

(according to Siemens application manual Magnetom Vision)
Vessel Flow velocity (cm/s)
Ascending aorta 50 – 100
Descending aorta 100
Aortic stenosis 150 – 500
Aortic valve insufficiency 150 – 200
Common carotid artery 60 – 80
Carotid artery stenosis 100 – 500
Middle cerebral artery 60
Basilar artery 40 – 50
Femoral artery 60 – 80
Popliteal artery 35 – 40
Vena cava 5 – 40
Portal vein 5 – 10

Fig. 29. Phase contrast study on a flow phantom with pulsating

flow. Left: venc = vmax ; Right: venc = vmax/2.
Due to vessel tortuosity and the time varying flow velocity, the sig-
nal intensity is higher when the flow sensitivity (venc) of the se-
quence is half the maximum flow velocity

Fig. 30. Phase contrast angiography before (left) and after (right) administration of a gadolinium contrast agent. Due to T1 shortening,
the signal-to-noise ratio increases, thereby improving the detection of small vessels

and 90 cm/sec) are encoded simultaneously but in
only one encoding direction. All three acquired
images are then combined to form a sum image.
In CE MRA, the signal-to-noise ratio of the
blood increases after contrast agent administra-
tion because the greater T1 shortening of the spins
leads to more signal. Phase contrast angiography,
unlike TOF angiography, can benefit from this ef-
fect without the penalty of increased background
signal. The additional signal afforded by the appli-
cation of a contrast agent (already at a dose of 4
ml) is particularly beneficial for the visualization
of small vessels with low flow velocity when work-
ing at lower field strength (Fig. 30).
As with TOF MRA, both 2D and 3D phase con-
trast MRA techniques are available. However, un-

Fig. 31. Phase contrast angiogram of the intracranial vessels
like TOF MRA, phase contrast MRA does not im-
pose any restrictions on image orientation, be-
cause the method is not dependent on inflow ef-
fects.
2D phase contrast sequences are well suited for
imaging vessels in a large volume. Because only
moving spins contribute to the measured signal,
the background signal is very effectively sup-
pressed. Even thick single slices of about 100 mm
can be imaged and the overall acquisition times
are relatively short (about 1-2 min). The result is a
projection image in which all the vessels in the ex-
cited slice volume are depicted (Fig. 31). A draw-
back, however, is that it is not possible to recon-
struct projections from another perspective.
With 3D phase contrast MRA datasets, as well
as with stacks of 2D phase contrast acquisitions,
MIP reconstructions can be acquired in any pro-
jection, in a manner similar to that performed in
3D TOF MRA. However, depending on the number
of partitions, the acquisition time for a 3D phase
Table 8. Application areas of MRA techniques
3D-TOF 2D-TOF
Intracranial:
- Arteries ***
- Veins * ***
Carotids ** **
Peripheral vessels **
*** method of choice; ** second-best alternative or for additional information; * working technique, but with sub-optimal results
TOF MRA: Time-of-Flight MRA; PC MRA: Phase Contrast MRA; CE MRA: Contrast Enhanced MRA
I.1 • Unenhanced MRA 21
contrast sequence is significantly greater than that
of a 3D TOF sequence, because four data volumes
need to be acquired rather than just one.
The possibility to combine ECG- or peripheral
pulse triggering with the 2D phase contrast se-
quence may allow the acquisition of time-resolved
images of pulsating blood flow in vessels. A series
of cine velocity images can be obtained that spans
the cardiac cycle. If an optimal depiction of pulsat-
ing blood flow is required rather than time resolu-
tion, it is advantageous to confine the acquisition
to the cardiac interval of maximum flow velocity.
As with TOF MRA, phase contrast techniques
can be combined with presaturation pulses in or-
der to eliminate unwanted vessels from the recon-
struction. However, because phase contrast se-
quences are independent of inflow effects, the sep-
aration of arteries and veins is not as successful as
in TOF-MRA.
Due to the flow encoding gradients, TE values
are prolonged compared with those employed in
TOF sequences, because the bipolar gradient puls-
es have to be applied within the TE interval. As a
result, phase contrast sequences are sensitive to
phase errors caused by, for example, susceptibility
effects or turbulent flow.
The sensitivity for flow is proportional to the
flow encoding gradient area (gradient amplitude *
gradient length). Therefore, the encoding of low
flow velocities requires longer TE values. Shorten-
ing of TE values can be achieved with stronger
gradients, by keeping the gradient area constant.
In this way, phase contrast MRA benefits from
stronger gradient systems.
Summary
In clinical practice, unenhanced MRA techniques
are mainly employed for imaging of the intracra-
nial vasculature. Although they are also applicable
to imaging of blood vessels in the neck, arm and
leg, CE MRA is now firmly established as the dom-
inating approach in the extracranial and periph-
eral regions. A summary of the different areas of
application for the different MRA techniques is
given in Table 8.
3D-PC 2D-PC Magnitude contrast CE MRA
* *
** * *
***
* ***
22 Magnetic Resonance Angiography
Table 9. Benefits and limitations of TOF and phase contrast MRA
TOF-MRA
Advantages Simple to implement, robust
High spatial resolution
Shorter acquisition time (in 3D)
Disadvantages Reduced sensitivity to slow flow
Restrictions to size and orientation
of the imaging volume
Short T1 tissue may be mistaken
for flowing blood
TOF and phase contrast MRA are based on
very different physical and technical mechanisms.
As a result, there are advantages and limitations to
both methods, which have to be taken into account
when deciding on the appropriate method for a
given vasculature territory (Table 9).
The TOF technique is robust and easy to imple-
ment. In the intracranial territory, it is established
as a screening method that allows good-quality vi-
sualization of the vascular anatomy. 3D TOF acqui-
sitions are particularly appropriate for imaging of
arterial flow. 3D TOF MRA provides high spatial
resolution with a significantly shorter acquisition
time than that required for 3D phase contrast
MRA. Problems due to saturation effects can part-
ly be avoided with the use of multi-slab and TONE
techniques. 2D TOF MRA is advantageous when
imaging veins, because of its higher sensitivity to
slow flow. On the other hand, it should be noted
that the quality of TOF MRA is strongly influenced
by flow velocity, the course of the vessels, and the
size and orientation of the imaging volume. The
use of very short TE values minimizes phase ef-
fects while poor background suppression can be
improved by applying magnetization transfer.
However, stationary tissue with short T1 may
sometimes be mistaken for flowing blood.
Unlike TOF MRA, phase contrast MRA is not
dependent on inflow effects. Therefore, the size and
orientation of the imaging volume can be chosen
arbitrarily. The method allows large sections of
vessels to be depicted almost without saturation ef-
fects. 2D phase contrast MRA requires only a short
acquisition time to generate a projection image of
one single thick slice while 3D phase contrast ac-
quisitions require a long time, since four datasets
have to be acquired to obtain full flow information
in all three orthogonal directions. Phase contrast
MRA is sensitive to slow flow rates and is applica-
ble mainly for imaging veins. Background suppres-
sion is excellent and short T1 tissues do not appear
Phase contrast MRA
No saturation effects
Excellent background suppression
Enables quantitative flow measurement
Prior knowledge about flow rates required
Very long acquisition times for 3D
techniques
Susceptible to phase errors
in the reconstructed angiograms. Because the sig-
nal intensity is directly proportional to the velocity
of local blood flow, phase contrast MRA permits
quantitative evaluations, such as flow quantifica-
tion or determination of flow direction. Problems
with phase contrast MRA may arise because prior
knowledge about blood flow rates is required in or-
der to avoid aliasing artifacts and flow voids.
References
1. Lenz G, Haacke E, Masaryk T, Laub GA (1988) In-
plane vascular imaging: pulse sequence design and
strategy. Radiology 166:875-882
2. Keller PJ, Drayer BP, Fram EK, Williams KD et al
(1989) MR angiography with two-dimensional ac-
quisition and three-dimensional display. Radiology
173:527-532
3. Laub GA, Kaiser WA (1988) MR angiography with
gradient motion refocusing. J Comput Assist Tomo-
gr 12:377-382
4. Parker GL, Yuan C, Blatter DD (1991) MR angiogra-
phy by multiple thin-slab 3D acquisitions. Magn Res
Med 17:434-451
5. Tkach J, Masaryk T, Ruggieri P et al (1992) Use of
Tilted Optimized Nonsaturating Excitation. SMRM
abstract 2:3905
6. Atkinson D, Brant-Zawadzki M, Gillan G et al (1994)
Improved MR angiography: Magnetization transfer
suppression with variable flip angles excitation and
increased resolution. Radiology 190:890-894
7. Edelman RR, Ahn SS, Chien D et al (1992) Improved
time-of-flight MR angiography of the brain with
magnetization transfer contrast. Radiology 184:395-
399
8. Axel L, Morton D (1987) MR flow imaging by veloc-
ity-compensated/uncompensated difference im-
ages. J. Comp Assist Tomogr 11:31-34
9. Dumoulin CL, Hart HR (1986) Magnetic Resonance
Angiography. Radiology 161:717-720
10. Dumoulin CL, Souza SP, Walker MF, Wagle W
(1989) Three-dimensional phase contrast angiogra-
phy. Magn Res Med 9:139-149