Joseph Mark

F. Polak, MD #{149} Richard L. Bajakian, MD R. Anderson, BS #{149} Magruder C. Donaldson,

Detection Comparison Sonography,

M

by using angiography as a standard in 23 consecutive patients (42 carotid bifurcations) to evaluate their utility

in determining the presence of carotid artery stenosis. MR angiography helped detect 50% or greater lumen diameter stenosis (sensitivity, 0.96; specificity, 0.64). Color Doppler sonography with 1.25 rn/sec peak systolic velocity as a threshold sensitivity of 0.96 and a specificity 0.71. Statistical analysis showed

had

a of

a

correlation between percentage of lumen diameter narrowing and the length of the zone of signal intensity loss with MR angiography (r = .69; P < .0001). A stronger relationship was obtained between angiographic narrowing and peak systolic velocity derived from color Doppler sonography (r = .80; P < .0001). Two-dirnensional time-of-flight MR angiography displayed as projection angiograms and combined with carotid artery sonography is a useful approach for helping detect and potentially grade the severity of stenoses of the carotid artery. Index terms: Carotid arteries, angiography, 172.124 #{149} Carotid arteries, MR. 172.1214 #{149} Carotid arteries, stenosis or obstruction, 172.721 Carotid arteries, US, 172.12984 #{149} Ultrasound Doppler

Radiology

From

‘

studies,

1992;

the

172.12984

182:35-40

Department

May

revision

dress (

13,

received

1991; July

reprint requests RSNA, 1992

H. O’Leary, MD A. Jolesz, MD

#{149} Ferenc

of Radiology

(J.F.P.,

revision 9; accepted

to J.F.P.

requested July

June 15. Ad-

Stenosis: Color Doppler

AGNETIC

tion of the findings obtained two-dimensional time-of-flight giography, as compared with and

with anangiog-

Doppler sonography with suspected carotid

MATERIALS

17;

Artery

resonance (MR) imaging renders anatomically correct representations of carotid artery bifurcations (1,2). A subset of this technique is two-dimensional time-offlight angiography, which shows increased signal intensity from flowing blood (3,4). These two-dimensional time-of-flight angiograms can be rapidly acquired. Projection angiograms similar in format to the traditional arteriograms can be created with computer processing (5). They can accurately depict the presence of flowing blood (1) but often show gaps or zones of signal intensity loss at the site of high-grade stenoses, where blood flow velocity is increased (2). There is great interest in whether MR angiography can be used to routinely screen the region of the carotid bifurcation for the presence of substantial stenoses. Duplex sonography is currently the noninvasive technique of choice for helping detect and grade the extent of atherosclerotic disease involving the carotid arteries (6-8). It is limited by the presence of large calcified plaques, its inability to demonstrate the intracranial portion of the carotid artery, and its heavy dependence on the skill of the sonographer (9,10). MR angiography does not have these limitations and can also be presented in a format that is acceptable to clinicians. We report on a prospective evalua-

raphy patients stenosis.

R.L.B., D.H.O., M.R.A., F.A.J.), Brigham and Women’s Hospital. 75 Francis St. Boston, MA 02115, and Department of Vascular Surgery, Harvard Medical School, Boston (MCD.). Received

MD

oflnternal Carotid of MR Angiography, and Arteriography’

Findings of two-dimensional time-offlight magnetic resonance (MR) angiography projection angiograms were prospectively compared with those of color Doppler sonography

(US),

#{149} Daniel

AND

in artery

METHODS

The results from a study of 23 consecutive patients with a high clinical suspicion of carotid artery disease were prospectively evaluated. The average age was 65.4 years ± 8.8 (range, 42-79 years). There were 13 men and 10 women. Five of the 23

patients terectomy,

had

previously undergone five had carotid bruits

endardetected

before coronary bypass surgery, 12 had a history of transient ischemic attack, and one had a history of recent stroke. Patients were screened for the presence of carotid stenosis

with

Doppler

sonography

per-

formed either in our own laboratory (13 cases) or in another laboratory (10 cases). Three of the Doppler examinations performed in our laboratory 21, 40, and 64 days before angiography were repeated on the day of carotid angiography, whereas the average time interval for the remaining 10 studies was 4.8 days (range, 1-13 days) before angiography. We repeated the 10 remaining studies, originally performed in other laboratories, on the same day as carotid angiography. Two of the MR angiograms were acquired the day before carotid arteriography was performed, whereas the remaining 21 angiograms were obtained the day of the carotid arteriograms. Two-dimensional

raphy Systems,

was

performed Milwaukee)

time-of-flight

angiog-

at 1.5 T (GE Medical after

an

MR

image

of

the brain was obtained to check for the presence of areas of infarction. A standard multiple plane coronal 600/20 (repetition time [TRJ msec/echo time ITEI msec) MR image (imaging time, 2 minutes 12 seconds) was obtained to localize the level of the common carotid bifurcations in the neck. Fifty-five consecutive axial MR images (50/li), with a section thickness of i.5 mm, a flip angle of 50#{176}, and bipolar gradients for flow compensation over a 20-cm field of view (imaging time, 6 minutes 31 seconds) were obtained. Saturation pulses were applied cephalad to the axial images to eliminate signal intensity from venous blood returning via the jugular veins. Image acquisition required less than 10 mmutes. Image reconstruction was performed with a Sun Sparcstation (Sun Microsystems, Mountain View, Calif) equipped with a TAAC-1 accelerator board. The images were transferred from the magnet by means of a high-speed Ethernet link with

Abbreviations: MIP = maximum intensity projection, TE = echo time, TR = repetition time.

35

proprietary

software

for

image

transfer,

data base management, and image reconstruction (GE Medical Systems). The images were reconstructed with the following steps: They were stacked in the random access memory of the TAAC board. A maximum intensity projection (MIP) algorithm was used to select the highest

pixel

intensity

at each

pixel

loca-

tion in the stack. The results were displayed as a transaxial view, showing a single collapsed image. With this image, a circular region of interest was selected, which was centered on the right carotid artery (Fig 1). This region of interest was then used and the MIP algorithm applied 30 times to create projection images corresponding

to 6#{176} increments

over

a 180#{176} rota-

tion. Serial display of these images gave a three-dimensional perspective. The last two steps were repeated for the left carotid bifurcation. The resulting images were sent back to the operator console for review and imaging. The transfer rate was approximately one image every 2 seconds. The presence of stenosis was determined by blind review of the MR projection angiograms and consensus among three readers (J.F.P., R.L.B., D.H.O.). Measurements were obtained from the image display

console.

A 50%

ing of lumen (Fig 2) or the 3) was

used

or greater

width of the carotid presence of a signal as an

indicator

the

internal

carotid

of hemody-

artery.

A

gate was positioned at this site, and Doppler spectra were obtained. At each site, the peak systolic and peak enddiastolic velocities were measured. A peak systolic velocity above i.25 m/sec was used as a determinant for the presence of hemodynamically substantial internal carotid artery disease ( 50% lumen diameter narrowing) (8,9). Angiography was performed with the standard femoral approach in 21 of the 23 patients; the brachial approach was required in two patients because of severe peripheral arterial disease. Intraarterial digital subtraction angiography was used in all patients and supplemented by cutscreen arteniography in three cases in which subtotal occlusion was suspected.

36

#{149} Radiology

1.

Diagram

tion angiograms is first acquired Small region

create

B

summarizes

the

of the carotid

angiograms

main

bifurcations.

over an 8-cm-long of interest is defined

projection

two

Series

segment on these

(B )

with

steps

of the images

the

MIP

in obtaining

of transaxial

and

gradient

displaying

refocused

neck centered on the carotid (curved arrow). Images are

MR

projec-

images bifurcation. then used

(A) to

algorithm.

narrow-

artery gap (Fig

namically substantial stenosis (4). Signal gap is a zone of signal loss that causes a visible gap in the course of the artery. This makes even low amplitude signals within the artery lumen indistinguishable from the surrounding background signals. The length of this zone of signal loss, when present, was measured from the projection angiograms. Real-time gray-scale imaging and color Doppler mapping were performed with 5and 7-MHz linear-array transducers (Acuson, Mountain View, Calif). The scanning protocol consisted of transverse imaging of the common carotid artery from its ongin to the level of the bifurcation. The transducer was then positioned parallel to the common carotid artery, and velocities were measured 2-4 cm proximal to the origin of the internal carotid artery. A color Doppler map was then used to detect the site of increased flow velocity within Doppler

A Figure

At least two selected projections were obtamed for each bifurcation. Of 46 bifurcations evaluated with MR angiography and color Doppler sonography, 42 were imaged with angiography. Selective angiography

was

not

used

with

four

carotid

bi-

furcations on the basis of the results of the MR angiographic and Doppler studies. Two patients had normal carotid arteries

on

sonograms

on

the

side

contralateral

to that with high-grade stenosis, and two patients had a long-term history of internal carotid occlusions that were documented with serial sonograms. These findings were confirmed with MR angiography. Among the 42 bifurcations, there were four instances of totally occluded internal carotid artery and one case of subtotal occlusion (99%) with a “string

sign.”

Because

antegrade

flow

is

substantially delayed, a prolonged injechon of contrast material and delayed imaging were required for the angiographic diagnosis of a string sign. Images were reinterpreted by an angiographer (D.H.O.) blind to the results of MR angiography and color Doppler sonography. The degree of stenosis was determined as the smallest residual lumen diameter narrowing seen on either of two views of the carotid bifurcation. Sensitivity and specificity were calculated, and 95% confidence intervals were recorded (11). Spearman rank correlation coefficients were used to compare the percentage of lumen arterial narrowing seen with angiography with MR angiographic size of signal intensity loss, as well as with peak systolic and peak end-diastolic velocities with Doppler waveform analysis. The relative importance of the correlations between these variables was assessed with the Friedman two-way analysis of vari-

ance by ranks. Comparisons groups were also made with tailed Student t test.

between the two-

RESULTS There were 14 internal carotid arteries with less than 50% lumen diameter narrowing seen at contrast-enhanced angiography and 23 arteries with at least 50% stenoses. In the 13 symptomatic patients, a stenosis of at least 50% was seen on the side with symptoms. There were four cases of total internal carotid artery occlusions, all of which were prospectively identified with sonography and MR angiography. The one case of subtotal occlusion was identified with MR angiography and Doppler sonography. MR imaging showed a faint string of signal intensity in the internal carotid artery, whereas Doppler sonography demonstrated a markedly decreased velocity (0.18 m/sec) in a narrowed flow lumen (Fig 4). The prevalence of 50% or greater stenosis or occlusion of the carotid bifurcations was 67% (28 of 42 bifurcations), and at least one of the two carotid arteries was involved in 91% (21 of 23) of the patients. The diagnostic sensitivity of MR angiography in helping detect greater than 50% stenosis was 0.96 (confidence intervals, 0.79 and 0.99; 22 of 23 arteries). The diagnosis of substantial stenosis was based on the visualization of a greater than 50% narrowing January

1992

P = not significant) or peak end-diastolic velocities (0.41 m/sec ± 0.29 vs 0.36 m/sec ± 0.15; P = not significant) when cases with a flow gap (n = 5) at MR angiography were compared with those without a flow gap (ii = 9). This was different in the patients with significant stenoses (Table 3). Peak systolic velocities were greater in patients with an MR angiographic flow gap (3.51 m/sec ± 0.96 vs 2.15 m/sec ± 1.03; P = .01). This was also true for the peak end-diastolic yelocities (1.21 m/sec ± 0.59 vs 0.70 m/sec ± 0.34; P = .02). The length of the zone of signal intensity loss in the internal carotid artery correlated with the percentage

a.

C.

Figure shows

2. (a) MR a moderately

origin

of the internal

arrows)

and

projection severe

a shallow

angiogram stenosis at the

carotid

artery

ulcer

(curved

(straight

arrow). helps confirm stenosis greater than 50% (peak systolic velocity, 1.69 m/sec). (c) Corresponding angiogram shows a narrowed lumen

(b) Corresponding

Doppler

sonogram

(straight arrows) at the origin of the internal carotid artery. Shallow ulceration is also clearly shown (curved arrow).

of stenosis seen with angiography (r = .69; P = .0001; n = 32). The correlations calculated between angiographic stenosis and peak systolic velocity (r = .80; P = .0001; n = 37), as well as end-diastolic velocity (r = .72; P = .0001; n = 37), were slightly better. A significant relationship existed between Doppler peak systolic velocities and angiographic estimates of the severity of carotid artery disease with the Friedman rank test, with a mean rank difference of 0.53, as compared with the least measurable difference of 0.85 (a = 0.05). This was not the case for other differences between variables (Table 4). The length of the zone of signal intensity loss at MR angiography also correlated with the peak systolic yelocity (r = .72; P = .0001; n = 32) and the end-diastolic velocity (r = .69; P = .0001; n = 32).

DISCUSSION

in five of the 22 instances diagnosis (Table 1). For 17 cases, the diagnosis was made on the basis of a measurable zone of signal loss in the internal carotid artery. The diagnostic specificity in 14 normal arteries ( < 50% narrowing) was 0.64 (confidence intervals, 0.39 and 0.84; nine of 14 arteries). The five cases with false-positive in the vessel of a positive the remaining

results showed a zone of signal intensity loss. One of these was at the site of a previous carotid endarterectomy. The length of the zone of signal intenVolume

182

#{149} Number

1

sity loss in these five averaged 0.34 cm (±0.25). A zone of signal intensity loss was seen in 14 of 15 cases with angiographic stenosis greater than 75% and in three of eight cases with stenosis of 50%-75%. In the 17 cases with stenoses greater than 50%, the measurable zone of signal intensity loss averaged 1.15 cm long ± 0.76. With a peak systolic velocity of 1.25 m/sec as a diagnostic threshold (Table 2), sonography helped detect 22 of 23 cases of 50% or greater stenosis (sensitivity, 0.96 [confidence intervals, 0.79 and 0.99]), whereas 10 of 14 normal ( < 50% narrowing) internal carotid arteries were properly classified (specificity, 0.71 [confidence intervals, 0.45 and 0.88]). When the i4 vessels with less than 50% stenosis were analyzed, there were no significant differences in either the peak systolic (1.31 m/sec ± 0.69 vs 0.98 m/sec ± 0.35;

We have shown that two-dimensional time-of-flight MR angiograms can be used to detect the presence of clinically substantial stenoses of the internal carotid artery by helping note the presence of either a nanrowed lumen or the development of a zone of signal intensity loss in the region of the stenosis. A narrowed lumen on MR projection angiograms was less common than the presence of a zone of signal intensity loss at the level of hemodynamically significant

(

50%) stenoses. It has long been recognized that a zone of signal intensity loss can develop distal to the site of a hemodynamically significant stenosis. This has been described both with time-offlight and phase-sensitive MR angiography. To our knowledge, this is the first attempt to correlate clinically the extent of this zone of signal intensity loss with both the angiographically Radiology

#{149} 37

determined rotid stenosis pler velocity

severity of internal and the increased that occurs at the

caDopsite of

stenosis. Our imaging protocol was implemented to be clinically applicable. The imaging sequence required 10 mmutes for signal acquisition. No sedation was needed or used. While this protocol was being evaluated, two patients were incapable of being imaged, one because of claustrophobia and one because of a pacemaker implant. The 23 remaining patients underwent diagnostically useful studies, with a success rate of 92% . Projection angiograms were used to display images in a form that was acceptable to both the radiologist and the referring physician. These projection angiograms were created with an MIP algorithm and were easily displayed in the same form as a standard angiogram. The high prevalence of diseased carotid arteries can be explained by our use of Doppler sonography as a first line for patient classification (10). An actual narrowing in the lumen was seen at MR angiography in five of 23 cases of internal carotid stenoses, whereas a zone of signal intensity loss was present in 17 of the 23 bifurcations with 50% or greater lumen diameter narrowing. The length of this zone correlated with the severity of stenosis and with the measured peak systolic and end-diastolic velocities of flowing blood. The zone of absent signal intensity or “flow gap” on the projection angiograms is caused by many factors (12). The signal intensity loss secondany to the phase dispersions occurring because of the increased velocity of blood and the presence of reversed flow downstream from the stenoses can be partly compensated for by shortening the TE of the sequence and obtaining thinner sections (13). The intravoxel dispersion linked to the turbulence that develops distal to stenoses cannot be completely eliminated. This accentuates the severity of the stenosis by making the MR flow lumen appear smaller or by actually causing a zone or gap without signal. Saturation effects, although less important in two-dimensional techniques, may similarly cause stenosis severity to be overestimated at MR flow angiography. These saturation effects cause the lumen of the normal carotid artery to look smaller, since the blood that flows close to the wall of the artery does so slowly and the signals it emits can be saturated in the same way as those from surrounding

38

#{149} Radiology

a.

C.

Figure

3.

shows

a zone

origin

of the internal arrows) and

straight

(a) MR

projection

of signal

intensity

angiogram

loss at the

carotid artery (solid no signal intensity at

the expected origin of the external carotid artery. Common carotid (curved arrow) and vertebral

(open

arrow)

arteries

are

clearly

visualized. (b) Corresponding Doppler spectrum acquired at this site helps confirm a zone of markedly increased velocity (3.87 m/sec). (c) Corresponding artenogram confirm the presence of a high-grade of the internal carotid artery (arrow) tal occlusion at the external carotid

helps stenosis and toartery.

soft

tissues. The application of the algorithm to axial gradient refocused images may mask the presence of lower intensity signals at or beyond

MIP

the stenosis. Signals from the static structures in the imaged section generate background noise after the application of the MIP algorithm. Therefore, the flow gap seen on the projection angiograms does not necessarily correspond to the zone of signal intensity void seen on the axial gradient refocused images. Ambiguity in selecting the highest signal intensity on a given projection may also be caused by the presence of smaller overlapping vascular channels. Errors can occur in the measurement of the length of the zone of signal intensity loss, which is related to the fact that the carotid artery does not necessarily stay perfectly perpen-

dicular to the transaxial imaging plane. In such cases, the oblique course of the vessel could account for the variability in the zones of signal dephasing resulting from a combination of partial volume and saturation effects (14). In our experience, grading the severity of internal carotid stenosis does not appear to be as reliable with MR angiography as with Doppler sonography. Thus, we recommend that the results of Doppler sonography be combined with those of MR angiography when grading the severity of a significant stenosis. phy is to be used,

the

If MR angiograsame TR and January

TE 1992

postoperative patients. It remains to be seen whether the turbulence observed in the internal carotid artery after endarterectomy also causes a false-positive finding at MR angiography. In the one instance in which a postoperative site had a 30% narrowing at angiography, MR angiography was graded as false positive and the Doppler peak systolic velocity was elevated at 2.29 m/sec. Because MR imaging is inherently sensitive to flow, important structural details such as plaque structure and size are lost. Although carotid plaque may be visualized with a more standard “black blood” imaging sequence (15), experiences with carotid artery sonography suggest that for stenoses greater than 50% lumen diameter nanrowing, Doppler velocity measurements tend to be more reliable than actual estimates of plaque size derived from high-resolution images. This observation is likely to apply to MR imaging as well and needs to be investigated more thoroughly. Although a two-dimensional time-

of-flight sequence was exclusively used in our protocol, it remains to be seen whether a three-dimensional approach might offer greater accuracy by eliminating

some

tual loss of signal from the oblique artery. Figure

The

MIP

algorithm

4. (a) MR angiogram reveals a diffusely narrowed lumen of the internal carotid artery (solid straight arrows). Signal intensity within the artery has decreased secondary to depressed flow velocities. Artifact (curved arrows) is secondary to patient movement during image acquisition. External carotid branches (open arrows) and vertebral artery (arrowhead) are clearly visualized. (b) Corresponding Doppler sonogram shows a markedly abnormal spec-

used for both techniques similar artifacts. In the

tnum with only low-amplitude antegnade blood flow in the artery. (c) Selective digital subtraclion carotid angiogram shows a high-grade stenosis of the internal carotid artery (straight arrow) with decreased filling of the more distal internal carotid artery (curved arrows). (d) Delayed image from a cut-film arteriogram shows a string sign and layering of contrast material in the dependent portion of the internal carotid artery (arrows).

selected

Table

I MR Imaging

of Internal

Artery

Stenosis

Severity

Stenosis

Severity

as a Function

of

Findings

Stenosis Severity with MR Angiography