Cerebrovascular magnetic resonance angiography: A critical verification

Transcription

1 Cerebrovascular magnetic resonance angiography: A critical verification G. E. Wesbey, MD, J. J. Bergan, MD, S. I. Moreland, MD, M. M. Sedwitz, MD, J. A. Bardin, MD, P. Schmalbrock, PhD, and J. Listerud, MD, PhD La fouu and Encinitus, Calif., Columbus, Ohio, and Philadelphia, Pa. Because simultaneous noninvasive noncontrast imaging of cervical and cerebral vasculature and brain is possible with magnetic resonance angiography (MRA) and imaging (MRI), the following study was undertaken from July 1990 to January One hundred twenty-eight patients were examined with General Electric 1.5 Tesla MRI systems. Axially acquired volumetric three-dimensional time-of-flight MRA with 0.7 mm 3 voxel size with regional maximum intensity projection after processing followed a two-dimensional time-of-flight localizing sequence. These two MRA sequences combined with spin-echo parenchymal brain MR/ were compared with duplex scans, contrast angiograms, and surgical findings. Blinded readings by a radiologist and vascular surgeon allowed comparison of grades of luminal diameter narrowing (normal, mild, moderate, severe, and occluded) seen on MRA to be compared with those of Doppler and contrast angiography. Excluding 12 nondiagnostically imaged internal carotid arteries (10 MRA) and limiting duplex correlation to within 5 days of the MRA examination allowed critical appraisal of 182 internal carotid arteries. Exact correlation of grade of stenosis was obtained by the radiologist in 136 (74.7%) of 182 arteries and the surgeon in 138 (75.8%) of 182 arteries. Spearman rank correlation analysis found rank correlation coefficients of 0.88 (p < 0.001) and 0.83 (p < 0.001), respectively, for the radiologist and vascular surgeon. Disagreement one category apart was found by the radiologist in 35 studies (I9.3%) and the surgeon in 28 studies (15.4%). Two or more grades of disagreement were found by the radiologist in 11 studies (6%) and the surgeon in 16 studies (8.8%). Contrast angiogram-mra agreement was found in 86% of 36 internal carotid arteries. The degree of stenosis detected by MRA was concordant with surgical findings in 39 of 40 patients. Thus MRA emerges as a useful and accurate method of obtaining cerebrovascular evaluation in clinical practice. (J VASC SUltG 1992;16: ) Stroke remains the third leading cause of death in this country, behind heart disease and cancer, and the most frequent cause of stroke is related to carotid artery ulcerative or stenotic atherosclerosis. Until now, radiographic contrast arteriography (XRA) has been the definfive method of evaluation of this condition. Its cost, risk, and lack of patient acceptability have limited consideration of its use in From the Departments of Radiology (Drs. Wesbey and Moreland) and Surgery (Drs. Bergan, Sedwitz, and Bardin), Scripps Memorial Hospitals, LaJolla and Encinitas; the Deparmaent of Radiology, (Dr. Schmalbrock), Ohio State University, Columbus; and the Department of Radiology (Dr. Listerud), Hospital of the University of Pennsylvania, Philadelphiia. Supported by a grant from the Vascular Center, Scripps Memorial Hospital Foundation, LaJolla, Calif. Presented at the Seventh Annual Meeting of the Western Vascular Society, Maui, Hawaii, Jan , Reprint requests: G. E. Wesbey, MD, Department of Radiology, Scripps Memorial Hospital, 9888 Genessee, LaJolla, CA /6/39930 symptom-free individuals who are at risk or otherwise suspected of having carotid artery stenoses. The importance of stroke as a problem has spawned much effort toward development of noninvasive methods of evaluation of carotid bifurcation. ~; Doppler ultrasonography has emerged as the screening procedure of choice and has been of clinical benefit in noninvasive evaluation of the extracranial cerebral vasculature. 3-6 Ultrasonography has negligible capabilities of imaging adult brain parenchyma. Although computed tomography can image the brain, it has assumed no place in evaluation of the extracranial cerebral vasculature. Therefore much recent effort has been expended in developing magnetic resonance imaging (MR/) techniques to study brain and blood vessels simultaneously. These have exploited time of flight (TOF) TM arid phase contrast to image the carotid and vertebral arteries. TOF techniques have become the mainstay in 619

2 620 Wesbey et al. Journal of VASCULAR SURGERY noninvasive magnetic resonance angiographic (MRA) evaluation of the extracranial carotid arteries, and both two-dimensional s12 and three-dimensional (3-D) 7'9'1a techniques have been employed with success. To date, all 3-D volumetric acquisition TOF MRA reports used in correlative studies of extracranial carotid artery disease have employed either sagittal or coronal planes of acquisition with voxel (individual volume element) sizes ranging from 1 to 2 mms. 7,9,13 Sectional acquisition two-dimensional (2-D) TOF MRA techniques have employed axial planes of section and voxel sizes ranging from 1.8 to 4.1 mm3. 8"12 This study was designed to evaluate the diagnostic accuracy ofa 3-D TOF cervical MRA with axial 0.7 mm 3 voxel acquisition by comparing magnetic resonance scans with Doppler ultrasound findings (sampling element 1.5 mm a) and XRA. A smaller voxel size in MRA should increase precision in interpretation of arterial anatomy. The MR findings were corroborated with operative findings at carotid endarterectomy (CEA) when patients underwent surgery. MATERIAL AND METHODS Between July 28, 1990, and January 2, 1992, 128 patients were referred for extracranial cerebrovascular MRA at Scripps Memorial Hospitals (La Jolla and Encinitas, Calif.). One of three General Electric 1.5 Tesla Signa MRI systems (General Electric Medical Systems, Milwaukee, Wis.) was used to study these patients. Two systems used the 4.6 General Electric Multisequence Vascular Package commercial software release version of MRA 2-D and 3-D TOF sequences that used digital radiofrequency-spoiled gradient echoes and fractional echo sampling of 70% with General Electric's homodyne version of partial Fourier reconstruction (n patients). The third system executed the General Electric 3.38 software platform with a research 3-D TOF sequence written by one author (P.S.) with nonspoiled gradient echoes, 75% fractional echoes, and full Fourier reconstruction (n patients). ~+ In 97 of these 128 patients a duplex ultrasonography study was performed within 5 days of the time of performance of the MRA study. Real-time gray-scale imaging and color Doppler mapping were performed with an ATL Ultramark-9 color system (Advanced Technology Laboratories, Bothell, Wash.) (n = 79 patients) or an Acuson color Doppler system (Acuson, Inc., Mountain View, Calif.) (n = 14 patients) with 5 MHz linear array transducers. The remaining four Doppler examinations were performed at outside institutions. Gray-scale axial imaging of the common carotid arteries from aortic arch origin to the bifurcation were followed by longitudinal color Doppler imaging to discern the location of the augmented flow velocity. The Doppler spectra were acquired at this high-velocity location in the internal carotid artery, as well as points in the distal internal carotid artery, and points in the proximal and distal common carotid artery, external carotid artery origin, and vertebral arteries. Peak systolic and peak enddiastolic velocities were measured at each site. Within this group of 97 patients in whom Doppler and MRA were performed within 5 days of one another, 12 patients also underwent XRA within 30 days of the MRA-Doppler studies. Nine of these 12 patients went on to CEA. An additional seven patients underwent CEA based on the MRA and Doppler examination without XRA. This provided a total of 16 patients undergoing 17 CEAs (one patient who underwent bilateral CEA) in this group of 97 patients, with operative correlation within 75 days of the MRA-Doppler studies. Although not undergoing a corroborative duplex examination within 5 days, a second group of patients was added to the evaluation pool. Twentyfive patients underwent 37 XRA carotid bifurcation studies within 30 days of the MRA examination. Within this additional patient pool were patients with Doppler examinations that were done greater than 5 days removed from the MRA examination. This group of patients increased the total number of patients by 31 and included 23 patients who underwent CEA within 75 days of the MRA examination. Eighteen of these 23 patients who underwent CEA in this second group underwent surgery based totally on noninvasive evaluation without XRA. Three of these patients who underwent CEA underwent surgery based on MRA alone. The Doppler-MRA 5-day group of 97 patients had an average age of 69.2 years (range 32 to 91 years). Thirty-two were outpatients and 65 were inpatients; 60 were men and 37 were women. Presenting indications were as follows: 10 patients were symptom free, 36 had transient ischemic attacks, 35 had stroke, and 14 had nonhemispheric symptoms. (History was unavailable for two patients.) MRA technique. All patients were studied on one of three General Electric 1.5 Tesla MKI systems with 10 mt/m gradients with body coil transmission and surface coil reception with a volume neck coil from Medical Advances (Milwaukee, Wis.). After a sagittal Tl-weighted locator image was performed, a series of low-resolution transaxial 2-D TOF MRA slices was obtained to locate the carotid bifurcation.

3 Volume 16 Number 4 October 1992 Cerebrovascular nurgnetic resonance angiography 621 After this, a high-resolution axial-body coil-transmit volume neck coil receive acquisition centered just above the carotid bifurcation was performed. The following technical parameters were held constant throughout the 18-month study: repetition time (TR), 37 msec; echo time (TE), 6.9 msec; flip angle, 20 degrees; 60 partitions of the 3-D volume; individual partition thickness, 0.9 mm; total axial 3-D slab thickness, 54 mm; field of view, 23 cm; matrix, 256 x 256; and one excitation. Flowcompensation gradients lz were activated along the readout and slice-select axes, and a superior saturation pulse Is with a 10 mm offset above the top of the 3-D imaging slab was used to suppress the signal from venous structures in the neck. This 3-D TOF MRA examination required 10 minutes for acquisition. Typically it spanned from the C3 to C6 vertebral bodies, depending on the location of the bifurcation. All patients subsequently underwent a routine double-echo spin-echo examination (TR range from 2200 to 3000 msec, with two echoes at TE 30 and TE 80 msec, flow-compensation gradients, and quadrature head coil). All 3-D TOF source axial partitions were then submitted for maximum-intensity projection algorithm after processing, with 30 projections obtained at 6-degree intervals starting from the left lateral projection. All 30 projections were filmed as hardcopy images. The projection processing was carried out as a background batch job on the Data General MV 7800 Eclipse computer (Data General Corp., Westboro, Mass.). To visualize better the anatomy, and to improve the signal/noise ratio of the resultant MRA projection image, a technologist-defined volume of interest was drawn around each carotid artery (circular 15- to 30-pixel region of interest around the carotid vessels selected from the maximal intensity projection [MIP] axial collapse), and a rectangular region of in-teres~ was used to define a limited MIP of the right and left vertebral arteries. Postprocessing was performed with IRMA (Image Reformation by Multiple Algorithms; General Electric Corporate Research and Development Center, Schenectady, N.Y.). The 30 projections of the right carotid artery, 30 projections of the left carotid artery, and 30 projections of the vertebral arteries were filmed as separate hard-copy images. This local MIP postprocessing minimized interference from surrounding stationary and vascular anatomy and dramatically diminished the processing time. In 58 patients, additional higher-resolution postprocessing of the carotid bifurcation was executed on the array processor with traced ray by array processor (TRAP) Fig. 1. Image shows normal 3-D TOF carotid bifurcation MRA obtained with axial plane of section and prospectively magnified selective volume-of-interest soft-thresholded maximal intensity projection postprocessing technique (TRAP). provided by one author (J.L.). This technique uses prospective volume bilinear/quadratic interpolation and magnification (user selected at 2.5-fold magnification) of the 3-D data set before execution of a soft-threshold maximum-intensity projection (Fig. 1). Conventional angiography. Contrast arteriography was performed through femoral artery catheterization. Of the 26 patients undergoing correlative examination with MRA and XRA, a total of 37 carotid arteries were studied, either by cut-film or intraarterial digital subtraction angiography. Duplex interpretation. The following criteria were employed for interpretation of duplex scans. Peak velocity measurements were aided by color Doppler imaging: 0% to 15% diameter reduction (normal) demonstrated minimal ( < 30 cm/sec) spectral broadening in the deceleration phase of systole, with peak systolic velocity less than 120 cm/sec; 16% to 49% (mild) luminal diameter reduction was defined as spectral broadening throughout systole (< 40 cm/sec), with peak systolic velocity less than 120 cm/sec; 50% to 79% (moderate) luminal diameter reduction by Doppler imaging was defined as peak systolic velocity greater than 120 cm/sec, with peak end diastolic velocity less than 130 cm/sec, with marked ( > 40 cm/sec) spectral broadening; and 80% to 99% (severe) luminal diameter narrowing by

4 622 Wesbey et al. Journal of VASCULAR SURGERY Table I. Vascular laboratory No. 1 validation (1990): Acuson Doppler-XRA correlation Doppler (%) 0%-15% 16%-49% 50%-79% 80%-99% 100% XRA Table II. Radiologist, MRA versus Doppler intemal carotid artery Dopplergrade MRA (%) 0%-15% 16%-49% 50%-79% 80%-99% 100% Doppler imaging was defined as peak systolic velocity greater than 120 cm/sec and peak end diastolic velocity greater than 130 cm/sec, with severe ( > 80 cm/sec) spectral broadening. Occlusion by Doppler was defined as no internal carotid artery Doppler signal and zero diastolic velocity in the common carotid artery. These criteria were used in all duplex scan evaluations, even the four examinations done outside our institutions, and follow the multicenter recommended criteria for standardized imaging and Doppler evaluation in duplex carotid ultrasonography. 19 Criteria for evaluating the Doppler ultrasound scans were uniform despite scan origin (see Duplex Interpretation). Validation of duplex scan interpretation was achieved during the time of this study by comparing scan results with carotid XRA in two parallel validation studies of our vascular laboratories. The Acuson Doppler system 1990 validation study (Table I) showed an 82% exact agreement of grade of luminal stenosis by Doppler criteria (see Duplex Interpretation) compared with carotid XRA grade in 16 patients studied (28 carotid arteries). A 100% agreement within one grade was achieved. The ATL color system validation study showed a 90% accuracy of grade of luminal stenosis by Doppler criteria (see Duplex Interpretation) compared with carotid XRA in 20 patients studied (37 carotid arteries) in 1990 (one two-grade error). MRA interpretation. The MRA examinations were reviewed retrospectively by a radiologist and vascular surgeon, blinded to the patient's clinical history, results of other examinations, and interpretation of other observers. A "forced-choice" evaluation of the distal common carotid artery, proximal extracranial internal carotid artery, proximal extracranial external carotid artery, and vertebral artery on both the right and left side of the neck was made. For the internal carotid artery specifically, the narrowed segment was compared with the normal distal internal carotid artery, similar to the approach taken by the trial method of the North American Symptomatic Carotid Endarterectomy Trial. 2 Estimates, not measurements, of luminal diameter stenosis were obtained and studies were reported as normal (0% to 15% stenosis), mild (16% to 49% stenosis), moderate (50% to 79% stenosis), severe (80% to 99% stenosis), or occluded. These readers had the following MRI information at their disposal: original 2-D TOF low-resolution axial slices; individual 3-D TOF axial partitions; and 30 local MIP projections from the 3-D TOF MRA data set individualized to the right carotid, left carotid, and both vertebral arteries. Also, the double echo-spin echo brain study was made available for interpretation. MRA evaluations by radiologist and vascular surgeon were then compared with those of duplex studies that used the same luminal diameter grading system. Correlation of results between techniques and observers was calculated by the Spearman rank correlation test.

5 Volume 16 Number 4 October 1992 Cerebrovascular magnetic resonance angiography Number of ICA Arterk~ Seg t~ Internal Carotid Artery Doppler - MRA Correlation 97 Pts Radiologist [] Surgeon Absolute Value Doppler Stenosis Grade Minus MRA Grade Fig. 2. Diagnostic accuracy of radiologist and vascular surgeon's MRA interpretations of internal carotid artery's (ICA) luminal caliber compared with that of duplex examination is depicted in bar graph. RESULTS Doppler-MRA correlation. Of the 97 patients who underwent Doppler and MRA examinations within 5 days of one another, 12 (6%) of the 194 internal carotid artery studies were judged to be technically unsatisfactory. Ten of these technically unsatisfactory examinations were MRA and two were Doppler. Metallic surgical clips were the cause of three of the 10 arteries rendered nondiagnostic on MRA; the bifurcation was missed in four others, and motion artifact degraded the other three. Duplex scan correlation was limited by calcific shadowing (n = 1) and unilateral examination (n = 1). The blinded interpretation of the MRA by the radiologist (Table II; Fig. 2) showed MRA and carotid duplex to be in exact agreement regarding the grade of internal carotid artery stenosis in 136 (74.7%) of 182 internal carotid arteries. In the 46 internal carotid arteries in which MRA and Doppler disagreed, the studies were only one category apart for 35 vessels (19.3%) and two or greater categories apart for 11 vessels (6%). The radiologist had nine two-grade discrepancies and two three-grade discrepancies. The Spearman rank correlation coefficient for this reviewer (r s = 0.88) was statistically significant (p < 0.001). The highest level of correlation was in the occluded category, with 94% of the vessels diagnosed correctly by the radiologist's interpretation of MRA. The greatest discrepancy between the two techniques was in the 16% to 49% category, with only 22% interpreted exactly. Thus 94% of the radiologist's MRA internal carotid artery category scores were correct or within one category grade of the Doppler interpretation. In the vascular surgeon's evaluation of the MRA of the internal carotid artery, compared with the carotid duplex study (Table III; Fig. 2), there was a 75.8% agreement (138 of 182 internal carotid arteries) (r s = 0.83; p < 0.001) (Table III). The vascular surgeon's MRA readings were within one category grade of the Doppler readings in 91.2% of the 182 internal carotid arteries. The category of highest agreement for MRA interpretation with the carotid duplex by the vascular surgeon was the normal (0% to 15%) category (93%; 100/108); the lowest percentage agreement was reached in the mild stenosis (22%; 4/18) category. Sixteen (8.8%) out of 182 internal carotid arteries were discrepant by at least two grades, with 14 two-grade discrepancies and two four-grade discrepancies. In the important category of internal carotid artery occlusion (Fig. 3), the radiologist was correct in 16 (94%) of the 17 cases of Doppler-documented internal carotid artery occlusion. In contrast, the

6 624 Wesbey et al. Journal of" VASCULAR SURGERY Fig. 3. Single image is seen of occluded internal carotid artery obtained by axially acquired 3-D TOF TRAP MRA projection. This lesion was confirmed by duplex examination. Table III. Vascular surgeon, MRA versus Doppler internal carotid artery Doppler grade MRA (%) 0%-15% 16%-49% 50%-79% 80%-99% 100% i : vascular surgeon correctly identified 11 (65%) of 17 internal carotid artery occlusions. In two of the six errors versus Doppler by the surgeon's MRA interpretation, the internal carotid artery was called normal. Two occlusions right at the internal carotid artery origin without a visible "stump" or cul-de-sac were overlooked because the first large ascending external carotid branch was mistakenly identified as a normal proximal internal carotid artery. In the eight patients in whom both the radiologist and vascular surgeon categorized a greater than or equal to two-grade discrepancy in the MRA internal carotid artery score versus the Doppler internal carotid artery score, four of the eight cases were upgrades of M1LA relative to Doppler and four of the eight cases were downgrades. In two of the eight cases the discrepancy between Doppler and MRA was resolved by another study. In one instance a Doppler evaluation found moderate narrowing of the internal carotid artery luminal diameter. However, both MRA interpreters scored the internal carotid artery as normal. An intraarterial digital subtraction angiographic study revealed a moderate stenosis at the internal carotid artery origin, confirmed at surgery. The 3-D TOF MRA image quality was degraded severely by patient motion artifact from daustrophobia. The radiologist interpreting the examination the day of the study called both bifurcations nondiagnostic, but both blinded readers chose to interpret the study. In another case the Doppler examination was categorized as mild (16% to 49%) narrowing of the right internal carotid artery origin. The MRA demonstrated a severe (80% to 99%) stenosis in the same location. This was confirmed at CEA 24 hours later. Doppler sampling and duplex imaging were markedly obscured by acoustic shadowing from dense calcific plaque. Agreement between the radiologist and vascular surgeon in interpreting the MRAs for all grades of the internal carotid artery was 138 of 182, or 76%

7 Volume 16 Number 4 October 1992 Cerebrovascular magnetic resonance angiography 625 Fig. 4. Image depicts occluded left vertebral artery on axially acquired 3-D TOF TRAP MRA projection, confirmed by Doppler. Slow flow in proximal basilar artery and distal right vertebral artery was seen in this 300-pound, 32-year-old man with acute posterior fossa neurologic deficit. (r s = 0.88;p < 0.001), with the highest correlation in interobserver agreement noted in the 80% to 99% category. The radiologist's interpretation of the MRA categories of common carotid artery stenosis, external carotid artery stenosis, and vertebral artery stenosis (Fig. 4) agreed exactly with the duplex examination in 96%, 79%, and 88% stenosis, respectively, with r s values of 0.73, 0.49, and 0.71 (all p < 0.001). The surgeon's MRA interpretation of the common carotid artery, external carotid artery, and vertebral arteries agreed with the Doppler examination in 96%, 76%, and 93% stenosis, with r s values of 0.75, 0.54, and 0.80 (all p < 0.001). Agreement of the two modalities within one category grade as assessed by the radiologist was 98%, 86%, and 95%; by the vascular surgeon, agreement within one grade for these three vessels was 98%, 91%, and 96%. For the common carotid artery the radiologist and surgeon agreed in 96% of the vessels; for the external carotid artery the two readers agreed in 79% of the arteries. For the vertebral arteries the two reviewers agreed in 95% of the patients. The following diagnoses were made on the parenchymal brain MRI examinations at the time of MRA: 41 patients with cerebral infarcts, 39 patients with small-vessel ischemic changes, 34 patients with normal parenchyma, and 3 patients with intracerebral hemorrhage. Conventional angiographic-m_ra comparison. For the XRA-MRA category of 25 patients and 37 carotid arteriograms, one of 37 vessels was nondiagnostic because of metal cfips. MRA agreement with XRA in the evaluation of the internal carotid artery was 86%, with a Spearman rank correlation coefficient r s of 0.95 (p < 0.001) (Fig. 5). Three MRA errors were overgrades and two were downgrades, with one two-grade error as mentioned previously caused by motion artifact (MRA was normal but XRA showed moderate stenosis, confirmed at CEA). No significant interobserver variability in the MRA interpretation was noted. Operative findings. The intraoperative CEA assessment of internal carotid artery stenosis was concordant with the MRA category in 39 of 40 patients (Fig. 6). The lone error occurred in the motion artifact-degraded MRA normal/xra moderate stenosis/cea moderate stenosis case mentioned twice previously. This study did not address the ability of MRA to detect ulceration. DISCUSSION The purpose of this study was to compare a previously unverified axial 3-D TOF technique of extracranial cerebrovascular MRA with carotid~ duplex angiography and XRA. This study did not address the issue of ulceration of the carotid bifurcation and dealt only with stenosis and occlusion. The high spatial resolution (0.7 mms voxels) and the extensive experience obtained with the axial plane of section and TRAP postprocessing in extracranial 3-D TOF MRA will allow this to be done in the future. Evolution of MRA technology has been so rapid that straightforward comparison of this modality to XRA or duplex ultrasonography of cerebral vasculature has been done by others, 913 with smaller numbers of cases than in this study. Mattle et al. 9 were able to collect 20 patients, Wilkerson et al.10 accumulated 13, Litt et al.xxgathered 50, Polak et al.a2 accrued 23, and Masaryk et al.13 garnered 38. MRA scans are so clearly physician and patient sufficient that angiography is ordered selectively even in surgical candidates. Because this practice also evolved during this study, it was necessary to compare MRA

8 626 Wesbey et al. Journal of VASCULAR SURGERY Fig. 5. A, Contrast carotid angiogram shows high-grade internal carotid artery stenosis and large ulcerative plaque. B, Axially acquired 3-D TOF TRAP MRA projection obtained in same patient faithfully reproduces angiographic findings in similar projection. with Doppler duplex examinations and with operative findings. Therefore a parallel confirmation of duplex scan interpretation and angiographic findings was accomplished, validating the duplex scan interpretations. Peak systolic and diastolic velocity examination at the site of maximal narrowing has associated well with angiography for plaque producing greater than 50% diameter stenosis. 19 This study has shown that both radiologist's and vascular surgeon's reading of MRA provides respectable results. This is true, although there is a learning curve for both physicians. With the extensive experience gained in our Radiology Depamiaent with 3-D TOF cerebrovascular MRA acquisition, processing, and interpretation, one would not expect a similar accuracy immediately with newcomers entering this arena. Problems affecting acquisition of acceptable MR.Is include slow arterial flow, patient motion, and metal clips or dental amalgam in the image field. Despite this, 95% of internal carotid artery segments in this patient experience were satisfactory. Of greater importance is the flow-gap phenomenon that occurs distal to critical stenosis of the artery. This produces a picture of pseudoocclusion. Because precise differ- entiation of occlusion from critical stenosis is crucial to patient care, the problem is important. It is partially resolved by acquiring 3-D TOF highresolution imaging and experience in performing the examinations and interpreting the images. Duplex Doppler examination was used in this study to compare accuracy of MRA with that of a study modality in common use. However, duplex scanning should not be considered competitive with MRA. Instead, it provides additional information. It does so in a noninvasive way and the combination of the two techniques becomes patient and physician tolerable. Clinically, situations arise in which MRA or Doppler duplex becomes the corroborative procedure of choice. For example, the Doppler provides inexpensive screening for suspected extracranial vascular disease. Magnetic resonance, in contradistinction, might be employed first for evaluation of brain parenchyma in patients with a neurologic deficit. When MRA is performed, duplex studies might or might not be added. Similarly, if severe carotid stenosis is found in a symptom-free patient, magnetic resonance could well show brain parenchymal deficits attributable to the carotid disease, as was demon-

9 Volume 16 Number 4 October 1992 Cerebrovascular magnetic resonance angiography 627 Fig. 6. A, Contrast arteriogram shows high-grade stenosis in proximal internal carotid artery. B, Axially acquired 3-D TOF TRAP MRA projection accurately displays atherosclerotic arterial anatomy. strated in this study when some studies showed unanticipated brain lesions. In many of our patients with clinically significant stenotic lesions of the internal carotid artery, the radiologist monitoring the examination performed an additional 3-D axial acquisition. This was done to position the 3-D slab so as to place the internal carotid artery stenosis in the caudal one third of the 3-D slab. There would be the least problem with slow-flow saturation and stenosis overestimation. Such tailored optimization of flow-related enhancement of carotid bifurcation stenoses is more difficult with commercially available MRA software with acquisitions employed in the sagittal or coronal planes because of software and hardware constraints. The major disadvantage of our 3-D TOF MRA protocol is the short vertical span (54 mm) projection image obtained. This might be annoying to clinicians and radiologists who are accustomed to highresolution arteriographic studies that provide continuous uninterrupted projectional display from aortic arch to middle cerebral artery genu. However, we partially meet this visual need with the 2-D TOF sequence that precedes the 3-D TOF MRA sequence. The MIP projections from this 2-D TOF examination provide a vertical lower-resolution (1.8 mm s voxels) arterial anatomy display for the clinician. This gives localization information for the radiologist to position the carotid bifurcation optimally for proper study of the patient's individual atherosclerotic le- sions and to screen for aortic arch great vessel disease. Carotid siphon atherosclerotic disease can be evaluated by intracranial 3-D TOF MRA as part of the brain examination. Just as magnetic resonance technology has evolved even during the course of this study, so has physician acceptance of the technique developed. This study addresses the primary problem of clinical importance: visualization of the carotid bifurcation to ascertain presence or absence of significant lesions in this location. The less common sites of carotid occlusive disease were not the focus of this article; neither is the histologic or biochemical study of the carotid bifurcation. As experience with TRAP postprocessing grew in this study, plaque complexity could be visualized, but this study does not address the frequency or validation of the interpretation of these findings. CONCLUSION MRA evaluation of extracranial cerebral vascularare done by a radiologist and vascular surgeon correlate with Doppler duplex studies and XRA studies with a reasonable degree of accuracy. Exclusion of redundant examinations with resultant cost saving is possible. Areas of furore study of clinical MRA should focus on the accuracy of ulcerative lesions, plaque accidents, and intraplaque hemorrhage and also the relationship of these lesions to parenchymal brain MR/ manifestations of cerebrovascular disease.

10 628 Wesbey et al. Journal of VASCULAR SURGERY We thank Patrick Hagan, PhD, Helen Engeseth, PhD, and David SherriU, PhD, of General Electric Medical Systems, Milwaukee, Wis., for their technical support of this project. We also thank the MRI, Doppler, and special procedure imaging technologists of Scripps Memorial Hospitals of La Jolla and Encinitas for their assistance. REFERENCES 1. Riccotta JJ, Holen J, Schenk E, et al. Is routine angiography necessary prior to carotid endarterectomy? J VASe SurtG 1984;1: Moore WS, Ziomeck S, Quinones-Bzaldrich WJ, Machleder HI, Busuteil RW, Baker JD. Can clinical evaluation and noninvasive testing substitute for arteriography in the evaluation of carotid artery disease? Ann Surg 1988;207: Steinke W, Kloetsch C, Hennerici M. Carotid artery disease assessed by color Doppler flow imaging: correlation with standard Doppler sonography and angiography. AJR Am J Roentgenol 1990;154: Glover JL, Bendick PH, Jackson VP, Becker GJ, Dilley RS, Holden RW. Duplex ultrasonography, digital subtraction angiography, and conventional angiography in assessing carotid atherosclerosis. Arch Surg 1984;119: Bornstein NM, Beloev ZG, Norris JW. The limitations of diagnosis of carotid occlusion by Doppler ultrasound. Ann Surg 1988;207: Thiele BL, Stranclness DE Jr. Duplex scanning and ultrasonic arteriography in the detection of carotid disease. In: Kempczinski R_F, Yao JST, eds. Practical noninvasive vascular diagnosis. 2nd ed. Chicago: Year Book, 1987: Masaryk TJ, Modic MT, Ruggieri PM, et al. Threedimensional (volume) gradient-echo imaging of the carotid bifurcation: preliminary clinical experience. Radiology 1989; 171: Keller PJ, Drayer BP, Fram EK, Williams KD, Dumoulin CL, Souza SP. MR angiography with two-dimensional acquisition and three-dimensional display. Radiology 1989;173: Mattle HI', Kent KC, Edelman RR, Atkinson DJ, Skillman JJ. Evaluation of the extracranial carotid arteries: correlation of magnetic resonance angiography, duplex ultrasonography, and conventional angiography. J VASe SURG 1991;13: Wilkerson DK, Keller I, Mezrich R, et al. The comparative evaluation of three-dimensional magnetic resonance for carotid artery disease. J VASc SURG 1991;14: Litt AW, Eidelman EM, Pinto RS, et al. Diagnosis of carotid artery stenosis: comparison of 2DFT time of flight MR angiography with contrast angiography in 50 patients. AJR 1991;156: Polak JF, Bajakian RL, O'Leary DH, Anderson MR, Donaldson MC, Jolesz FA. Detection of internal carotid artery stenosis: comparison of MR angiography, color Doppler sonography, and arteriography. Radiology 1992;182: Masaryk AM, Ross JR, DiCello MC, Modic MT, Paranandi L, Masaryk TJ. 3DFT MR angiography of the carotid bifurcation: potential and limitations as a screening examination. Radiology 1991;179: Schmalbrock P, Yuan C, Chakeres DW, Kohli J, Pelc NJ. Volume MR angiography: methods to achieve very short echo times. Radiology 1990;175: Dumoulin CL, Souza SP, Walker MF, Wagle W. Threedimensional phase contrast angiography. Magn Reson Med 1989;9: Bendel P, Buonocore E, Bockisch A, Besozzi MC. Blood flow in the carotid arteries: quantification by using phase-sensitive MR imaging. AJR 1989;152: Haacke EM, Lenz G. Improving MR image quality in the presence of motion by using rephasing gradients. AJR 1987;148: Felmlee JP, Ehman R.L. Spatial presaturation: a method for suppressing flow artifacts and improving depiction of vascular anatomy in MR imaging. Radiology 1987;164: Bluth EI, Stavros AT, Marich KW, Wetzner SM, Aufrichtig D, Baker JD. Carotid duplex sonography: a multicenter recommendation for standardized imaging and Doppler criteria. Radiographics 1988;8: North American Symptomatic Carotid Endarterectomy Trial. Methods, patient characteristics, and progress. Stroke 1991; 22: Submitted Feb. 13, 1992; accepted June 2, DISCUSSION Dr. Thomas McNamara (Los Angeles, Calif.). Since the use of magnetic resonance (MR) scans, unlike computerized tomographic scans, blood vessels can be seen without the use of contrast material; there has been great interest in developing clinically reliable MR angiography. This would be a particularly attractive capability for exttacranial cerebrovascular evaluation in that it is without the small but definite risks of angiography. It is not as operator dependent or as adversely affected by calcium as ulttasonography, and it can, as recently shown a few moments ago, yield useful information about the target organ by obtaining a brain scan during the same examination. Considerable progress has been made in both hardware and software to overcome the difficulties that are inherent in obtaining MR images of blood vessels that are reliably obtained and approach the accuracy of either ultrasonography or angiography. This advancement has been associated with increasing optimism and enthusiasm for the technique. That is not unwarranted, but it may lead the unwary to rely on the technique too heavily too soon. A brief and a simple review of the problems inherent in the current MR imaging methods may be useful because only a few of those in attendance have any reason to be familiar with MR principles. The presence of air or bone adjacent to flowing blood

11 Volume 16 Number 4 October 1992 Cerebrovascular magnetic resonance angiography 629 can cause considerable artifacts. This is manifest by the high number of vertebral artery examinations that were inadequate. Blood flow is normally laminar but, as you know, not uniform. The so-caued boundary layer at the interface between blood and intima flows more slowly than blood in the center, and current imaging methods often lose that layer. Thus the vessel is depicted as narrower than it really is. To image flowing blood, the parameters for MR imaging, including the time to echo, time for repetition, and flip angle, have been modified significantly in various software programs. As implied by previous remarks it is not completely standardized. Thus images from one institution may vary if more than one operator directs the examinations, and variance from institution to institution can be considerable. This makes reading of MR images more difficult. An article in this month's issue of Rad/0/0gy documented that shortening the time to echo reduces the size of the flow gap, and they called for uniformity of technique parameters. I am afraid that will be sometime in coming. The most commonly used method for carotid imaging at this time is called 2-D TOF, which has been presented briefly. It obtains a stack of axial scans that are theoretically at right angles to the flowing blood and then projects them so as to look like an angiogram. The cervical carotid artery is actually a relatively straight artery in many instances, but increasing obliquity results in an artifactual loss of the signal intensity. This method relies on blood flowing straight into and out of that plane of section. That is obviously not the case when the flow is turbulent as a result of a stenosis or when it is flowing horizontally within that plane of section because of tortuosity or kinking. The flow must also fall within certain speed limits or the signal is lost. The 2-D TOF, therefore, frequently demonstrates loss of signal, and it is known to overestimate the degree of disease. It is most accurate for demonstrating anatomy of a normal or mildly stenotic vessel. The 3-D TOF imaging does not require that the flow be perpendicular to the sampling plane. Thus tortuosity, kinks, and turbulence are not sources of artifact. However, inherent in the physics of this type of sampling is the requirement to image only a small volume of tissue. Otherwise, resolution caused by the saturation suffers dramatically, because the protons in the flowing blood are bombarded by radiofrequency signals so many times that they begin to wobble about and do not give a reliable return signal. The length of the vessel depicted was only 54 mm at a time. Thus very accurate positioning of the sampling volume is required. Another method of imaging that is called black-blood imaging is beyond our time limits for discussion today. Resolution for ulceration is clearly an important issue but unfortunately not part of this study. In reading the Results Portion of this article, I was struck by the need to depict in a table format the various comparisons. You have done a wonderful job of comparing many things (e.g., comparisons between observer accuracy, MR and ultrasonography, MR and arteriography, and MR and carotid endarterectomy). I found it difficult to keep track of all the numbers as one progressed from paragraph to paragraph and would suggest that they be put in table format. In an effort to validate or test the accuracy of the method, you have looked at many vessels in the same patient. I think that that is a very good idea, but the most important vessel in these examinations is the internal carotid artery. Although you state that a 93% agreement between MR and ultrasonography was present, that actually was true only if one not only included the correlations in which both the MR and sonogram were read as being in the same degree of stenosis but also if one included as accurate when one was off by one grade. It would seem that it would be more appropriate to focus on the incidence of complete agreement. That incidence was approximately 75%. I understand the situation that the presenter was in, but I was nonetheless disappointed that in this MR angiographic correlative study only 25 of the 127 patients also underwent arteriography. You note that in the majority of patients with significant internal carotid artery stenosis the radiologist monitoring the examination performed at least a second 3-D examination in which the sampling volume or slab was positioned so as to have the stenosis in the caudal third of that slab. How much did that improve the statistics and what would have been the accuracy if only the first slab had been used? Can we rely on a technician to perform that task and monitor the examination? In view of your experience, what would you propose as a practical algorithm for selecting which patients undergo ultrasonography, which patients undergo MR, which patients undergo angiography, and which patients would go from one examination to another? Under what circumstances will MR be sufficient to proceed directly to carotid endarterectomy? Three patients were treated in that manner. Could you explain why, and what are the reasons we should apply in our own practices? We believe that we understand the reasons MRA overestimates the severity of stenosis, but what were the causes for the underestimates? Are they inherent in the method or were they observer error? Would you please discuss underestimates of both stenosis and occlusion separately? The clinical limitation of 3-D imaging for carotid disease is the short vertical length of the vessel demon strated, as mentioned. That makes precise placement of the sampling volume critical. It seemed that some of the errors were the result of incorrect localization of that carotid bifurcation. How can one, in a practical clinical setting, prevent that error and does that require that a radiologist supervise each examination? Do you foresee that the technology for 3-D imaging

12 630 Wesbey et al. Journal of VASCULAR SURGERY will improve such that a greater length of the carotid artery will be able to be visualized? What are your thoughts, if you have been exposed to this information, on the highresolution 3-D technique currently being used by Tjeruda and Norman at the University of California, San Francisco, in which, instead of the 256 by 256 matrix used in your own and other methods of 3-D MR imaging, they were using 512 by 512? Dr. Wesley S. Moore (Los Angeles, Calif.). Unfortunately, the world that we live in is one in which there is a tendency to incept in perhaps a premature manner new technology in both a diagnostic and therapeutic vein. I would be interested in a show of hands of those of you who are seeing a reasonably large number of patients either as a referral or, in fact, being operated on based on MR alone. That is refreshing. I was afraid I was going to see a sea of hands, but that has happened to a large degree in our own community and it is something that we need to guard against. Dr. George E. Wesbey. I want to emphasize that this has been an 18-month project under my supervision, and obviously one of the take-home messages is that this is the state of the art as it is applied at Scripps. Absolutely, I would not expect these results to be duplicated in a community hospital setting or maybe a university setting because one of the points we make is that this technique of high-resolution 3-D (supplementing a 2-D TOF) angiography with 0.7 mm 3 voxel sizes is the smallest voxel size reported to date for clinical verification of extracranial carotid MR angiography. The previous studies have been on the order of 1 to 1.5 mm 3 voxel size, and the question vascular surgeons should take home to their radiologists who are performing 3-D time of flight is, "What voxel size are you using and is it competitive with these numbers?" Let me begin by emphasizing the first issue about positioning of the slab and radiologist monitoring. Yes, that is essential and, no, it should not be placed in the hands of a technologist at this point. In terms of in how many of our cases of stenosis were additional smaller slabs placed so that the stenosis was at the caudal one third of the imaging volume, I do not have the exact numbers in front of me, but I would say roughly in 25 to 30 patients additional 3-D slabs were placed after initial radiologist review. I have no idea what the accuracy would be of this technique if we went with the initial 60 partition examination alone, but I am sure it would be less. Specifically what this additional positioning does is provide more optimal positioning of the stenosis in the caudal one third of the slab, allowing better flow-related enhancement and essentially improved conversion of what we call a flow gap to what you are accustomed to seeing on conventional arteriography, a string sign. As far as the proposed algorithm goes, it is our belief at this point that the Doppler examination should serve as the initial screening procedure in symptom-free individuals (e.g., those patients who harbor carotid bruits and perhaps patients who will be undergoing coronary artery bypass grafting). We believe MR has an important role in the initial evaluation of patients who have some kind of neurologic deficit who do not have a contraindication to MR imaging. We can provide one-stop comprehensive cerebrovascular parenchymal and vascular evaluation. I do want to emphasize that the safety contraindications are very important: pacemakers, intracranial aneurysm clips, and metal in the eye. Please remember that. Do not send us patients with a pacemaker. Those patients are excluded from MR imaging. As far as the role for x-ray angiography, it is my belief that if the 2-D TOF flight images of the arch origin of the great vessels raise any question of a proximal arch origin lesion, conventional arch aortography obviously needs to be performed. If there is any question about fibromuscular displasia in a young patient, I believe that is another clear need for conventional arteriography. A third would be as we screen the carotid siphon with intracranial MR angiography, the siphon turns out to be a very difficult area for MR angiography because of the turbulent flow. Dr. Saloner has developed some nice techniques with appropriate sagittal slab positioning to improve the accuracy, but I would say at this point that we do not have any large experience at all with siphon stenoses; I believe conventional arteriography would be needed to confirm or deny the presence or absence of an MR angiography-suspected siphon stenosis. Finally, we have had at least two very interesting cases of common carotid artery occlusion in which the issues of transfacial collateral flows are oriented horizontally from right to left from the external carotid circulation. The depiction of external carotid collateral flow is still a job for conventional angiography and not MR angiography. So if you have a patient with a common carotid artery occlusion and there are external collaterals, the internal carotid artery can still be opened by the external transfacial collaterals. Depiction of those collaterals is impossible, in my opinion, based on the small numbers we have had with MR angiography. So that would be yet another indication for x-ray angiography. With regard to the question of operating on the basis of MR angiography alone, the three cases in which we have operated on the basis of MR angiography alone are a very recent experience (in the last 3 weeks at the end of this 18-month study, performed by Dr. Bardin). He has had the most experience in our group operating on the basis of MR and Doppler alone (18 endarterectomies). He feels comfortable in selected individuals in whom there is a very high-quality combined 2-D and 3-D TOF MR angiographic examination. Needless to say, at this point I still would feel most comfortable with a combination of Doppler and MR angiography in technically optimal examinations. I think that is another key point, that you have to develop a sense of aesthetics for what is a quality 3-D TOF MR examina-

13 Volume 16 Number 4 October 1992 Cerebrovascular magnetic resonance angiography 631 tion and what is a dog meat examination. That is really important. In terms of the underestimations, I think the most important category was in the occlusion category. We had 17 cases of Doppler-delineated occlusions. The radiologist scored 16 of the 17 occlusions as occlusion by MR angiography; the vasoalar surgeon scored 11 of 17. Relating to the statistics in the previous article, 19 cases were graded as 80% to 99% stenosis by Doppler. The radiologist called three of those 19 occlusions, and the vascular surgeon called one of those 19 occlusions. I think the second to the final question was, "Am I optimistic that 3-D TOF axially acquired techniques will be able to be lengthened beyond the 54 mm vertical dimension that we use at Scripps?" I would say quite honestly no. That is based on the TOF principles as Dr. Saloner mentioned. The longer the spins reside in the 3-D imaging slab, the more they get saturated; I think perhaps the other important take-home technical point from our article in addition to the voxel size is the axial plane of acquisition. The vertical dimension of a sagittal or coronal 3-D slab is on the order of 180 to 230 mm. The so-called dwell time of the spins is determined by the length of the 3-D slab. If the stenosis is at the top of a 3-D slab with a 230 mm slab height, there may be more tendency to overestimate the stenosis than if it is at the bottom of a much smaller, more focused slab. On the other hand, the trade-off of this technique is supervision, and I want to emphasize supervision and analysis. We put nearly as much time in monitoring these examinations as we do with contrast arteriography. The last question regarding high resolution, we will be getting the 512 package from General Electric in a couple of weeks, and we are very excited about that. We have begun a study of ulceration, and needless to say I think that will be of immense importance. The final point I want to make is the critical importance of the postprocessing. These just do not come out of the computer as is, and the key point, as Mark mentioned, is the traced ray by array processor postprocessing developed by Dr. John Listerud at Penn. This is a unique program. Penn is the only other one in the world with this program. The bottom line of traced ray by array processor is prospective bilinear and quadratic volume interpolation of the 3-D data with soft thresholding creating markedly improved arterial delineation of the extracranial carotid arteries. Dr. David A. Saloner (San Francisco, Calif.). You brought up the question of whether the size of the carotid artery that is covered can be increased or it is reasonable to expect it to be larger than 54 mm. In our experience we do have the option of reducing the flip angle, increasing the field of view, and getting a sagittal slab. This, again, depends somewhat on the artery you have with a neck coil receiver. On the neck coil you may not be able to do this. Once you identify the location of the lesion, it is easy to go in and do the high-resolution narrow-slab acquisition. So I think we will get a better overview of the vessel with those techniques. With respect to the supervision of the study, it is definitely very important that radiologists have input, but we have the experience in our institution where we have a number of technologists working on the system, three or four of them, who are able to carry out carotid examinations without supervision. This was not the case in the study we reported. They have been trained to do these carotid examinations on volunteers with very little input from our radiologists. They are able to screen the carotid artery, identify the location of interest, place the slabs appropriately, and postprocess the data, perhaps not as efficiently as our most highly trained technologists, but they still do an adequate job. So we believe that will be a trend in the future. Dr. Gregory L. Moneta (Portland, Ore.). Studies obtained at our institution are, in my experience, not of the same quality as those shown in your presentation. Is this technique easy to learn? Are radiologists in practice just applying the available software without really understanding what they are doing? How well educated are general radiologists in the application of these techniques? Dr. Wesbey. Speaking for myself, I feel I have needed every minute of my 10 years in MR angiography to apply myself to answer patient care problems as posed. To answer your question, you are alluding to a big problem and that is continuing education for community radiologists for MR angiography. Quite frankly I think it must be overwhelming to private practice community radiologists to learn how to interpret the flow physics of MR angiography. So, rather than give a general answer to the question, I will acknowledge that it is definitely a problem the magnitude of which I do not know. I think what you as vascular surgeons need to do is to go back to your radiologist and ask what MR angiography courses they have been attending. Have they been, for instance, to the University of California, San Francisco, MR course? Whatever the bottom line is, this is not something you pick up on the fly from reading a stack of articles on a plane, and it requires a lot of self-education, as well as validation within your department. Dr. Eugene F. Bernstein (La Jolla, Calif.). I wondered if you could help us as surgeons as we bring you our patients and our problems. Are there clues for us to look for in the quality of MR reading dealing with artifacts or other things? Vascular surgeons usually read their own films, as you know, and obviously we will be more comfortable if we can learn to read MR angiograms. Could you talk to us about artifacts and other technical things we need to learn about as we become accustomed to this method? Dr. Wesbey. First, the Bible for me is the article by Anderson et al. (Am J Roentgenol 1990;154:623-9). We made John Bergan review this outstanding pictorial atlas before every reading session. Again, post-processing introduces artifacts that are after