Joseph
E. Heiserman,
MD,
Intracranial Evaluation Time-of-Flight
PhD
Burton
P. Drayer,
#{149}
MD
R
studies
ECENT
documented of two- and time-of-flight mag-
resonance
(MR)
the evaluation carotid artery
Cerebral
blood
vessels,
or obstruction, 17.721 #{149} Magnetic (MR), three-dimensional, 17.7214 resonance (MR), vascular studies,
Radiology
stenosis resonance
Magnetic
#{149}
17.7214
185:667-673
1992;
I From
the
Barrow
Neurological
Institute,
St
print v
requests
RSNA,
to B.P.D.
1992
angiography
potential
lateral
to the
blood
supply
distributions
of the
evaluation toms
the
include
sympof
or infarction. METHODS
Twenty-nine patients, all of whom had focal neurologic deficits of varying duration, were studied to exclude vascular stenosis or occlusion. All patients underwent MR imaging of the brain, three-dimensional
time-of-flight
MR
of the intracranial vessels, tional angiography. The tween MR and conventional varied from 0 to 29 days The study group included female patients, with an years (range, 4-78 years). The MR examinations Medical dimensional,
were
performed
(Signa;
Milwaukee).
Fourier
angiography
and conventime interval beangiography (mean, 5 days). 13 male and 16 average age of 56
system
Systems,
GE
The
transform,
three-
time-of-
flight MR angiographic images were obtained with a gradient-echo pulse sequence with “spoiling” of residual transverse magnetization (5) known as spoiled gradient-recalled acquisition in the steady
angle used. used, partition
(6). This a repetition
sequence time
was implemented of 50 msec and
a flip
of 20#{176}. No presaturation pulses were Typically, a 256 x 256 matrix was with a field of view of 20 cm and a thickness of 0.7 mm, resulting in
a 0.78 x 0.78 x 0.7-mm
voxel.
were
thickness 13 minutes
of 4.5 cm in an imaging time of 40 seconds. Six of 29 examina-
(21%)
were
acquired
within
Sixty-four
partitions
tions
acquired
192 matrix and a 19-cm field of view, with a corresponding decrease in imaging time. All examinations were obtained with first tion-
with
moment flow compensation and frequency-encoding
a section
a 256
x
on secaxes. The
echo time varied during the course of the study from 10 msec to 4.9 msec due to technical improvements: 16 of 29 (55%) were
obtained
msec, two (7%) (38%) at 10 msec.
The
evaluation
MD
Occlusion:
ischemia
AND
K. Fram,
4.7
the intracranial arteries. In this article, we explore the usefulness of threedimensional time-of-flight MR angiography of the intracranial vasculature in the assessment of patients with symptoms of cerebral ischemia
MATERIALS
Evan
#{149}
images
full
with
cerebral
ideally
for colvascular
brain,
of the patient
suggesting
would
in
of narrowing of the bifurcation (1-4). Ow-
ing to the extensive
state with
Joseph’s Hospital and Medical Center, 350 W Thomas Rd, Phoenix, AZ 85013. Received April 13, 1992; revision requested May 28; revision received July 7; accepted July 13. Address re-
and
efficacy
three-dimensional netic
PhD
have
the clinical
on a 1.5-T imaging terms:
Keller,
Vascular Stenosis with Three-dimensional MR Angiography’
To assess the usefulness of magnetic resonance (MR) angiography in the characterization of intracranial arterial stenosis and occlusion, a threedimensional time-of-flight method was compared with conventional angiography in 214 vessels in 29 patients. Studies were independently interpreted by two neuroradiologists who scored each vessel as normal, narrowed, or occluded. Overall, 97% of normal vessels and 100% of occlusions were correctly graded. Sixtyone percent of stenoses were graded correctly; the remainder were graded as normal. The portions of the intracranial vessels near the skull base and especially the paracavernous and supraclinoid segments of the internal carotid arteries were areas of frequent over- and underestimation of stenosis due to the presence of dephasing artifacts. In patients with stenosis or occlusion, MR angiography also provided information regarding the presence of collateral flow in the circle of Willis. When used in conjunclion with MR imaging of the brain and MR angiography of the extracranial carotid arteries, intracranial MR angiography allows a more complete evaluation of the patient with symptoms of cerebral ischemia or infarction. Index
PaulJ.
#{149}
this
64 axial
way
were
source
with
echo
time
at 5.3 msec,
and
images
obtained
postprocessed
by
of
11 in
means
of
a maximum-intensity-pixel projection algorithm (7) to produce multiple projeclions rotated about the section axis. The anterior and posterior circulations were processed separately, with regions of interest that excluded orbital and scalp fat. In each case, 10 projection images at 18#{176} increments were obtained, as well as one axial “collapse” projection image. Selective conventional angiography of the carotid and vertebral artery distnbutions was performed in all cases. A combination of magnification screen-film and digital-subtraction images (512 x 512 ma-
trix) were
obtained.
Oblique
views
and
views with cross compression were not obtained routinely and were not included in this study. The MR and conventional angiograms for each patient were reviewed retrospeclively and independently at separate reading sessions by two neuroradiologists (B.P.D., J.E.H.). The MR angiograms included the projection images and the source images for each patient. For the conventional angiographic studies, anteroposterior and lateral images from the selective carotid and vertebral injections were supplied. The readers were blinded to all clinical information. A standardized scoring scheme was employed with three grades: normal, narrowed (50% diameter narrowing or greater), and occluded. Six vessels were scored bilaterally: internal carotid artery, the horizontal (A1) and postcommunicating (A2) segments of the anterior cerebral artery, the horizontal (M1) and distal (M2) middle cerebral artery, and the posterior cerebral artery. The yessels in the circle of Willis were evaluated for variants such as hypoplasia and fetal origin of the posterior cerebral artery. From these data, tables were constructed that compared the scoring of a
667
Figure ample image.
1. Normal MR angiograms. (a) Exof source image. (b) Axial projection (c, d) Anteroposterior-projection im-
ages segmented from data for posterior circulation, respectively. time = 4.9 msec in a-d.
anterior and Echo
given vessel by reader and by technique. For purposes of this study, left- and rightside vessels were considered equivalent, and results were pooled. The percentage agreement and the Spearman rank correlation coefficient T5 (8) were calculated to compare the results for each reader by technique and also to compare the readers for each technique. Part of the observed agreement between readers and tech-
niques due this
in measurements
of this kind
is
to chance alone. To take account of effect, the K statistic (9), which mea-
sures
agreement
beyond
that
expected
on
the basis of chance, was also computed. To examine the effect of echo time on the accuracy of MR angiography, the subgroup of examinations obtained with echo time of 4.9 msec was analyzed separately and findings were compared with those obtained with echo time of 10 msec, by means of the correlation coefficient and K
a.
b.
statistic.
RESULTS All MR and conventional angiowere judged acceptable by both readers, although the quality one MR angiogram was considered grams
suboptimal
and
small
communicating scored on
due
to motion
vessels
such
this
of
artifact,
as the
arteries examination.
were
C.
posterior not A normal
MR angiogram is illustrated in Figure 1. The paracavernous and supraclinoid segments of the internal carotid arteries basilar
are
are visualized, as is the distal artery. The ophthalmic arteries
identified
occasionally,
although
fat signal intensity arising from the orbits on these Ti-weighted images limits visualization of vessels in this
area. The superior are usually seen. tions tenor
cerebellar The proximal
of the anterior, cerebral arteries
variable
portions
ments
of these
The
middle, and posare well seen;
of the vessels
A1 segment
arteries por-
distal are
seg-
visualized.
of the anterior
ce-
rebral artery was absent or hypoplastic in 16 of 58 (28%) of both examinalions (Fig 2). Of the 56 posterior communicating arteries scored, 43
(77%)
were
grams and angiograms. the posterior
identified
on MR angio-
39 (70%) on conventional About half (23 of 43) of communicating arteries seen at MR angiography were present only on the source images and not on the projection images.
668
Radiology
#{149}
d.
A total of eight fetal origins of the posterior cerebral artery, with a hypoplastic or aplastic P1 segment of the ipsilateral posterior cerebral artery, were observed at both MR and conventional angiography. Duplication
of a superior present
in two
cerebellar of 29 cases
artery (7%)
was and
was
apparent
in
both
ventional angiograms. cies of these common
served
in this study
MR
and
con-
The frequenvariants ob-
are comparable
to
the
accepted values (iO). A total of 214 vessels were assessed for acquired narrowing or occlusion by means of both projection and
December
1992
the
internal
carotid
arteries
and
the
middle cerebral arteries, and scoring for these vessels for each of the readers is summarized in Tables 2 and 3. For this case, the four correlation coefficients
two (P
between
readers
basis
C.
Figure 2. Hypoplastic A1 segment of the right anterior cerebral artery as an example mal variant of the circle of Willis. (a, b) Conventional right and left carotid angiograms (c) MR angiogram in the anteroposterior projection (echo time = 4.9 msec).
of a norand
incorrectly
Evaluation at Conventional Angiography Reader
by Technique Evaluation
Normal
for Each
at MR Angiography Narrowed
Ocduded
collateral
ages was
1 188
9
0
2
9
0
Occluded Reader 2 Normal Narrowed Occluded
0
0
6
filling
this
186 9 0
5 8 0
was
Good
correlation
is apparent,
Spearman rank correlation cients rs of .73 and .65 and
Volume
185
Number
#{149}
3
with coeffi-
P value
less
middle,
Findings angiograms
0 0 6 and
posterior
cerebral
was
present
artery
by the posteartery. In one
clearly
in the
of the
apparent
on
the
were
than .005 for readers 1 and 2, respectively. Overall agreement in findings at MR and conventional angiography was seen in 203 of 214 (95%) vessels for reader 1 (K .71) and in 200 (93%) for reader 2 (K .64). Most of the abnormalities were observed to involve
sured
with
of MR
with
echo
compared
ings in those obtained of 10 msec. Improved arteries.
pathway cases.
subgroup
obtained
conventional of internal carotid and anterior, distal branches are not scored.
source images (in the case of MR angiography). The overall agreement in scoring between techniques for each reader is summarized in Table 1.
were
as occluded.
(Fig 4). No collateral identified in these
of 4.9 msec
are numbers occluded vessels,
readers
MR projection angiograms; in the other, this collateral pathway was apparent only on the source images. In two cases of middle cerebral artery occlusion, infarction was present and was apparent on the source im-
Reader
Normal Narrowed
Note-Numbers For proximally
the
graded
middle cerebral rior communicating
1 of 214 Vessels
data,
In four cases in which an occluded internal carotid artery was present, flow was present in the affected anterior and middle cerebral arteries, and the pattern of collateral filling was evaluated (Fig 3). In all four cases, collateral filling was present of the antenor cerebral artery via the anterior communicating artery. In two cases,
case,
of Evaluation
of these
in 103 of 114 (90%) MR angiograms (K .64) and in 102 (89%) conventional angiograms (K .62). In both cases, no occlusion was incorrectly graded as normal or narrowed, and no normal or narrowed vessel was
Comparison
the
to 1.0
in agreement
b.
Table
0.71
rowing was both over- and underestimated at MR angiography compared with conventional angiography in a number of cases. The agreement between readers for each technique for the internal carotid and middle cerebral arteries is summarized in Table 4, with correlation coefficients rs of .74 and .86 and P value less than .005 for both MR and conventional angiography. On the
4.
a.
for
from
.005). The six occlusions present identified by both readers in all Most normal vessels were coridentified; however, vessel nar-