Evaluation of the Aortic Root by MRI
Insights From Patients With Homozygous Familial Hypercholesterolemia
Background—In homozygous familial hypercholesterolemia (HFH), the aortic root is prone to develop atherosclerotic plaque at an early age. However, the aortic wall and plaque have not yet been assessed in this condition by MRI. We evaluated the aortic root by use of MRI in 17 HFH patients and 12 normal control subjects in a prospective, blinded, controlled study.
Methods and Results—Morphological assessment of the aortic root was done with spin-echo and gradient-echo MRI scanning. Comparisons were made with a number of measures of disease severity, including cholesterol-year score, calcium score on electron-beam CT (EBCT), and size of Achilles tendon xanthomas. Atherosclerotic plaque, visible on fat-suppressed images but never on water-suppressed images, was present in 9 HFH patients (53%). Supravalvular aortic stenosis was present in 7 patients with HFH (41%). Maximum supravalvular aortic wall thickness was significantly greater and OD and lumen cross-sectional area (CSA) were smaller in patients than in control subjects (P=0.006, 0.0005, and 0.06, respectively). Maximum wall thickness was associated with a greater calcium score on electron-beam CT (P=0.02). Although the cumulative exposure of the aortic root to cholesterol (the cholesterol-year score) was significantly correlated with the Achilles tendon CSA and vascular calcification, this score did not correlate with the wall thickness or aortic CSA.
Conclusions—This study not only demonstrates the utility of MRI for detecting and characterizing aortic root atherosclerotic plaque and supravalvular aortic stenosis in HFH patients but also suggests that the LDL receptor plays a direct or indirect role in aortic mural development and vascular growth.
Homozygous familial hypercholesterolemia is an uncommon inherited form of hypercholesterolemia that serves as a model for our understanding of the development of atherosclerosis. Patients with HFH have extremely high serum cholesterol levels and may develop advanced atherosclerotic plaque before 10 years of age.1 The plaque formation can occur at unusual sites, including the ascending aorta and around the coronary ostia.2 These atheromata can interfere with aortic valve function3 and cause patients to present with angina, myocardial infarction, and even sudden death.4
Familial hypercholesterolemia is an autosomal dominant disease characterized by elevated LDL in the blood. The primary defect is a mutation in the gene for the receptor for plasma LDL.2 More than 150 different mutations are known to exist. The gene defect causes a deficiency in the number of functioning LDL receptors. Because of this deficiency, LDL removal from the blood is impeded and excess LDL accumulates in scavenger cells, producing xanthomas and atheromas.2 Phenotypic homozygotes, although rare (numbering 1 in 1 million persons), are the most severely affected because their cells take up little or no LDL. Treatment includes cholesterol-lowering drugs, plasmapheresis, coronary artery bypass surgery, and in some cases, liver transplantation.5
Because atherosclerosis is such a frequent finding in HFH, patients with HFH serve as a useful model for the understanding of atherosclerosis. New evidence suggests that atheromata in the aortic root may be a heretofore unrecognized cause of cerebral ischemic events.6 Thus, the high frequency of aortic root lesions in HFH patients lends another rationale for the development of techniques to detect these lesions.
In this study, we evaluated the aortic root in HFH patients and normal volunteers by MRI to detect the presence and location of plaque, to assess morphological (wall thickness, CSA) and hemodynamic (valvular regurgitation, turbulence) features, and to correlate these findings with serum cholesterol and duration of disease in the patients. Findings were also correlated with the presence of aortic root calcification on EBCT and ankle xanthomas demonstrated by CT. We investigated several MRI pulse sequences to evaluate which ones were most efficacious for detecting plaque. The study was done prospectively, with image analysis done by investigators blinded to the clinical history. To the best of our knowledge, this is the first reported MRI study of the aortic root lesions in HFH and the first to suggest that the LDL receptor may be important for vascular growth.
Seventeen HFH patients and 12 normal, healthy volunteers were studied. Demographic data are summarized in Table 1⇓. The HFH patients represent a consecutive series referred to our center for treatment. We estimate that our institution (National Institutes of Health, Bethesda, Md) sees one third to one half of all patients with HFH in the United States. Other characteristics of some of these HFH patients have been reported previously.7 8 Two patients and 2 normal volunteers were scanned more than once. For these subjects, only data from the first complete MR imaging session were used.
Although there was no statistical difference between the age distributions of normal volunteers and HFH patients (P=0.32), there were 6 patients <18 years of age and no normal volunteers under this age. We were unable to recruit pediatric volunteers because we were not approved to do so through our Institutional Review Board. Written informed consent was obtained for all subjects, and the study was done after Institutional Review Board approval. Anthropometric data (height, weight, blood pressure, and heart rate) were obtained from all patients and from the majority of the volunteers (Table 1⇓).
The diagnosis of HFH was based on plasma lipoprotein analysis and family history.7 The patients were subject to a number of cardiovascular complications (Table 2⇓). Patients and control subjects were not similar with regard to history of hypertension, smoking history, or diabetes, and we did not attempt to control for these factors.
The severity and duration of hypercholesterolemia in the patients were determined by use of the “cholesterol-year score.” This score is computed by multiplying the initial serum cholesterol value (in mg/dL) by the age at diagnosis to compute the pretreatment score and then adding the annually determined posttreatment total cholesterol values.7
All MRI scanning was done on a 1.5-T Signa MR unit (General Electric Medical Systems). Imaging included spin-echo and gradient-echo sequences.
Oblique axial spin-echo images of the aortic root and proximal ascending aorta transverse to the lumen were obtained with a phased-array surface coil. The imaging sequence used ECG gating, fat suppression, and respiratory compensation, with 5- or 7-mm slice thickness, no interslice gap, 18-cm FOV, 256×160 matrix, NEX=2, TR=1 RR interval, and TE=12 ms. Saturation bands were placed above, below, anterior to, and posterior to the aortic root. The superior and inferior saturation bands suppressed MR signal from flowing blood, which otherwise could lead to image artifacts. Nonsuppressed and water-suppressed sequences were also done. The number of images obtained depended on heart rate but ranged from 5 to 11 images.
Oblique axial gradient-echo cine images perpendicular to the long axis of the aorta were obtained at 16 phases of the cardiac cycle at each of 3 contiguous levels in the supravalvular aortic root of 12 patients and 9 normal control subjects. Gradient-echo images were added to the imaging protocol after the beginning of the study. The imaging parameters were TR=99 ms, TE=9 ms with fractional echo, 7-mm slice thickness and 7-mm gap, 18-cm FOV, flip angle 20°, 256×160 matrix, 16-kHz bandwidth, and NEX=2. Both respiratory and flow compensation were used. Gradient-echo images on 3 HFH patients that were of unacceptable quality because of technical failure were discarded from the analysis.
EBCT was done as part of a different arm of our institution’s protocol on HFH. Transaxial scans through the thorax were done with ECG gating on an Imatron scanner with settings of 130 kV, 620 mA, 100-ms exposure, slice thickness 3 mm, and FOV=15 cm. Calcium scores, which represent the total amount of calcium in the coronary arteries and thoracic aorta, were derived from region-of-interest measurements.9 10
Achilles tendon CT was done to assess for the presence of xanthomas.11 The ankles were scanned without the use of intravenous contrast in the flexed position with toes pointing straight up. Scans were obtained with 120 kVp, 280 mA, 1-second exposure, contiguous slice thickness 10 mm, and FOV as small as possible to include both tendons. The CSAs of the tendons were measured bilaterally at the maximal diameter by hand tracing.7
Measurements of aortic wall thickness were made on the first image of the supravalvular ascending aorta above the sinus of Valsalva. Additional images through the ascending aorta were evaluated for focal abnormalities; the number of useful additional images depended on the length of the ascending aorta and its curvature, because the images were all parallel and contiguous. If a focal area of thickening was present, it was measured and its location recorded. The aorta appears round on oblique axial cross-sectional images and can be represented as a clock face, with 12 o’clock and 3 o’clock representing the anterior and left walls, respectively. Measurements were made at the 12 positions around this clock face to assess regional variations in wall thickness. A measurement could be made only if both sides of the vessel wall were delineated. Occasionally, the outer wall could not be detected if there was insufficient contrast with mediastinal tissue (eg, myocardium, adjacent vessels, mediastinal soft tissues). The position of the outer wall was revealed best by suppression of the adjacent fat. A visual approach was used to determine wall position. These measurements were made by a single observer blinded to the clinical data by use of random code numbers assigned to each MRI study. Average and maximum wall thicknesses on the first supravalvular section and regional wall thickness averaged on the first 2 supravalvular sections at each of the 12 positions around the clock face were determined. Focal or diffuse areas of wall thickening were designated as areas of plaque formation.
We tested the hypothesis that the aortic root is small in patients with HFH. The CSA of the aortic root was measured just superior to the sinus of Valsalva. This location was chosen because of the known tendency of HFH patients to develop supravalvular aortic stenosis. Measurements were made from the gradient-echo cine images at end diastole (first phase after the R wave) by tracing the outer contour of the bright signal from flowing blood. Each measurement of CSA was made 3 times, and the mean was computed. To test the effect of body habitus on the results, CSA was normalized to BSA or weight. BSA was computed from weight and height.12
The presence of supravalvular aortic stenosis was determined from the fat-suppressed images. The 2 images just above the sinus of Valsalva were examined, and supravalvular aortic stenosis was diagnosed if the transverse inner dimension of the aorta on the more inferior slice (just above the sinus) was <90% of that from the more superior slice. The choice of 90% as a cutoff was made on the basis of our observations of the normal control subjects.
The transverse IDs and ODs of the aortic root on the first supravalvular section were measured from the fat-suppressed images by a single observer as another determination of aortic root hypoplasia. The transverse dimension was chosen because the outer wall of the aortic root could be reliably determined there.
The presence on gradient-echo images of turbulence or jets indicative of abnormal flow states (including aortic stenosis or regurgitation) was recorded.
Unpaired 2-tailed t tests were performed to determine the significance of comparisons between data from the control group and the patients, with P<0.05 considered significant. Pearson’s correlation (r value) was computed to test significance of cholesterol-year score and calcium score with MRI morphometry of wall thickness, diameter, and CSA. Intraobserver and interobserver variabilities were expressed as a coefficient of variation computed from 3 measurements by a single observer and measurements from 3 different observers of a representative random subset of the data from 6 patients.13
A typical image of the ascending aorta is shown in both a normal volunteer and a patient with HFH (Figure 1⇓). Note the thickening of the wall of the aorta and narrowing of the lumen in the patient study. The patient and the normal volunteer are the same age and sex, yet the diameter of the patient’s ascending aorta is one half that of the normal volunteer. The BSA for the patient was 29% less than that of the normal volunteer, however (1.5 versus 2.1 m2).
The fat-suppressed images were best for measuring wall thickness (Figures 2⇓ and 3⇓). The nonsuppressed and the water-suppressed sequences were less useful. On nonsuppressed images, the mediastinal fat overlapped a portion of the vessel wall, obscuring it and making it difficult to measure its thickness, because of chemical shift artifact. On water-suppressed images, only the mediastinal fat remained in the images; the aortic wall and plaques became invisible. The fat-suppressed sequence was best because it avoided the chemical shift artifact and did not suppress the plaques.
Plaques were visible on MRI studies from 9 patients and appeared as irregular focal or diffuse wall thickening (Figure 3⇑). All involved the left wall of the aorta and were similar in signal intensity to and indistinguishable from the wall. Comparison of nonsuppressed and fat-suppressed images showed no obvious fat within the plaques. In 2 patients with known calcific plaque, there was reduced signal intensity within the plaque on MRI (Figure 4⇓).
The average wall thickness measured on the first supravalvular section was 3.3±1.2 mm (range, 1.8 to 6.9 mm) in the patients and 2.3±0.5 mm (range, 1.7 to 3.3 mm) in the normal control subjects (P=0.004). The maximum wall thickness measured on the first supravalvular section was also greater in the patients than in the normal control subjects (Table 1⇓ and Figure 5⇓). The latter measurement would tend to reflect the presence of a focal lesion or plaque.
Analysis of regional wall thickness showed that the wall thickness of the patients’ aortas was uniformly greater on average than that of the control subjects at all positions around the clock face. The greatest statistically significant difference was on the left side, at the 3 o’clock position (P=0.001), and the least significant difference was at the 11 o’clock position (P=0.2). When patients who had plaque were compared with those without plaque, the wall thicknesses in the supravalvular aortic root were also uniformly greater.
Seven HFH patients had evidence of supravalvular aortic stenosis. Of these 7 patients, 6 also had evidence of plaque in the aortic root.
Measurements of supravalvular aortic CSA showed that the aortic cross section is smaller in patients than in control subjects, although this result was not statistically significant (P=0.06) (Table 1⇑). Six patients had CSA <400 mm2, but only 1 control subject did. When these data were normalized to BSA or body weight, the correlation was poorer and the normalized CSA values were larger in patients than in control subjects (P=0.91 and 0.20, respectively). Normalization to BSA or body weight heavily weighted data from children <12 years old. In addition, HFH patients were shorter, weighed less, and had lower BSA than normal control subjects. Because of these characteristics of the HFH patients, we did a subgroup analysis comparing the 8 adult HFH patients ≥18 years of age with the normal control subjects. Only patients who had had gradient-echo MR images were included in this subgroup. For the subgroup, CSA, CSA corrected for body weight, and CSA corrected for BSA were also not statistically significant (P>0.25), although the normalized values were smaller than those for the normal control subjects.
Both the IDs and ODs of the supravalvular aortic root were significantly smaller in patients than in normal control subjects (P<0.0005; Figure 6⇓). On average, the IDs and ODs were 5.5 and 5.8 mm smaller in patients than in normal subjects, respectively. The disparity in diameter between patients and control subjects held true if adult patients alone were considered (P=0.03) but not for BSA-corrected diameter measurements for adults (P<0.09).
Turbulence was identified at the level of the aortic valve in 2 patients and no control subjects. Turbulent flow was identified on the cine images by the presence of flow void that entirely filled the lumen during peak systole. Very high flow rates can also produce this flow void. One patient with turbulence had an aortic root graft and aortic valve prosthesis in place. The second patient had aortic valvular stenosis and marked thickening of the aortic valve leaflets. Jets of aortic insufficiency were identified in 5 patients and no control subjects.
There was a mild correlation between aortic OD corrected for BSA and cholesterol-year scores, although the correlation was not statistically significant (P=0.06, r=−0.46; Figure 7⇓). Wall thickness (both maximum and average) and lumen CSA were not associated with cholesterol-year scores (P≥0.23). When MRI studies on HFH patients were segregated into 2 groups, those with (n=9) and those without (n=8) plaque on MRI, there was still no statistically significant association with cholesterol-year score (plaque: 17 300±10 100; range, 3100 to 32 300; without plaque: 10 100±7000; range, 4200 to 23 200; P=0.11).
There was a strong correlation between aorta calcium score and cholesterol-years (r=0.80, P=0.0002). The combined Achilles tendon CSA was also strongly correlated with cholesterol-years (r=0.88, P=0.000003).
For the patients, no associations were found with age or sex for either maximum wall thickness or aorta CSA (P≥0.27). The combined Achilles tendon CSA measurements were not correlated with either maximum wall thickness, aorta CSA, or average wall thickness (P≥0.07). Aortic calcium score was correlated with maximum wall thickness (r=0.59, P=0.02) but with neither aortic root CSA nor average wall thickness (P≥0.09). The calcium score/wall thickness correlation was heavily weighted by 3 outliers. Neither tendon size nor calcium scores were associated with the presence of plaque (P≥0.09).
The coefficients of variability for intraobserver and interobserver measurements were 10% (95% CI, 8% to 12%) and 24% (20% to 28%), respectively, reflecting a mean SD of wall thickness of 0.3 and 0.8 mm, respectively.
Our results demonstrate that the wall thickness of the ascending aorta is increased in patients with HFH. Although this was partly because of the presence of focal plaques, there may be a generalized component of wall thickening that is consistent with the known medial hyperplasia present in this disorder. Generalized wall thickening was most apparent in the youngest patients, for unknown reasons (Figure 3⇑). Because plaques can regress with treatment, serial MRI of the aortic root may be useful in HFH patients to determine prognosis and response to therapy.
We found a significant correlation between maximum wall thickness measured on MRI and the aortic root calcium score determined from EBCT. This finding is not surprising, because plaques can cause an apparent focal thickening of the wall (detectable on MRI) and plaques can calcify (detectable on EBCT).
A correlation between severity and duration of hypercholesterolemia and volume of plaque calcification on EBCT has previously been demonstrated in the setting of HFH.8 We confirmed the correlation between disease severity (cholesterol-years) and calcium score in a larger patient population, and we detected plaques in the aortic root, but we were unable to confirm a statistically significant correlation between disease severity and either presence of plaque or thickness of the aortic wall.
We found that plaques tended to be on the left side, although the wall thickness was greater at all locations along the aortic wall. The reason for this discrepancy is unknown. Studies have shown that atherosclerotic plaque and intimal thickening tend to form in areas of recirculation and low wall shear stress,14 15 possibly because of differential gene expression.16 Geometric factors, such as the curvature of the aortic arch, are also known to be important.17 18 Perhaps the aortic annular abnormalities and supravalvular aortic stenosis modify the blood flow velocity profiles into a pathological configuration that predisposes to plaque formation in these locations.
Supravalvular aortic stenosis was present in 41% of patients. Overall, the CSA of the supravalvular aortic root was 23% smaller, the ID was 22% smaller, and the OD was 19% smaller in the patients than in the control subjects, although these differences disappeared when the measurements were corrected for body weight or surface area. Plaques were found in 86% of the patients with supravalvular aortic stenosis, although the plaques were not always present on the section just above the sinus of Valsalva, on which the diagnosis of stenosis was made. We did not measure the outer CSA of the aorta. Therefore, the smaller inner CSA in the patients could be due to wall thickening encroaching on the lumen or to a small aorta. The lumen diameter data, however, suggest that the aorta is small.
Supravalvular aortic stenosis is a characteristic feature of HFH.4 19 20 The cause of the supravalvular aortic stenosis in HFH is unknown. One possible explanation is that it is due to altered growth of the vessel wall, because the atherosclerosis of HFH occurs at such an early age (E.C.J. et al, unpublished observations). High LDL concentrations might repress the expression of genes related to aortic growth. The lack of correlation of the lumen CSA and cholesterol-year scores would seem to refute these hypotheses. However, it may be the levels of LDL cholesterol during childhood that are important, rather than cumulative cholesterol-years. Alternatively, LDL receptor expression may be important for cellular growth and proliferation. Rapidly dividing cells express high levels of LDL receptors.21 22 23 In the absence of normal LDL receptors, the growth of the ascending aorta may be arrested in these patients.
We were able to see aortic root thickening consistent with atherosclerotic plaques on MRI of 9 patients with HFH. In the evaluation of clinical atherosclerotic disease, the role of MRI has been primarily to detect stenoses and occlusions of the arteries rather than to visualize the plaque itself. However, atherosclerotic plaques can also be visualized on MRI,24 and the first in vivo MRI studies of atherosclerotic plaques in patients are beginning to appear.25 26
In our study, plaques were seen only on fat-suppressed images and never on water-suppressed images. This finding is consistent with results of other studies. For example, 1 study showed that plaques are well seen on fat-suppressed images and better discriminated from periadventitial fat than on conventional spin-echo images.27 In that study, the plaques became invisible after water suppression. This was attributed to the negligible amount of isotropic (liquidlike) signal from immobilized lesion lipids. Spectroscopy showed the lipid resonance to be broad and ill-defined. A later study of excised human arteries with atherosclerotic plaque also found that fat suppression did not significantly change the appearance of the plaque.28 Because the spectral width of the fat-suppression pulse is relatively narrow, the lipids in plaque (some of which have a broad peak) are not suppressed, but the triglycerides in periadventitial fat (which have a relatively narrow peak) are suppressed.
In our study, the plaques had a relatively bland appearance, with little apparent structure. Calcific plaques were poorly seen at MRI, and the size of plaque calcifications on EBCT corresponded poorly with the MRI appearance. Our results suggest that MR shows wall thickening better when calcifications are absent in contrast to CT.
This study has several limitations. The spin-echo images were all parallel to one another. Therefore, some of the images were oblique to the true cross section of the aorta. This could cause an artifactual thickening of the aortic wall as the aorta curves relative to the plane of section. We compensated for this limitation by orienting the slices correctly for the supravalvular aortic root and using only 3 to 5 of the 9 images in the spin-echo data set, thus spanning only 2 to 4 cm of aorta. Images in which the aorta was grossly oblique to the imaging plane were discarded from further analysis, but subtle biases in the data due to this effect could still occur. Another limitation was that, although gated to the ECG, each image was obtained at a different phase of the cardiac cycle. Thus, some images of the aortic root were obtained during systole and others during diastole. However, both the patients and the control subjects were scanned by the same technique. Also, one would not expect the appearance of plaque to change significantly between systole and diastole. These limitations were due to constraints imposed by limited MRI scanning time and the duration of scanning tolerable to patients in 1 sitting. Gradient-echo imaging was begun after the beginning of the study, and 3 HFH patients had gradient-echo images of unacceptable quality. Therefore, only a subset (two thirds) of the subjects had CSA measurements. This limitation is reflected in the poorer correlations for the CSA data. Diameter measurements were not subject to this limitation, however. There were significant anthropometric (age, BSA) differences between patient and control groups, and thus, they were not perfectly matched. This was unavoidable because HFH strikes patients when they are young, and we were not allowed to scan normal children. We controlled for these differences by performing subgroup analyses.
To date, detection of aortic root and arch atherosclerosis is not a standard part of a cardiovascular workup and is often an incidental finding. For example, calcified plaques are often detected on CT scans of the chest in elderly patients. However, a recent study has shown that increased aortic wall thickness is a risk factor for recurrent ischemic stroke.6 We have shown that MRI scanning with a standard clinical MRI scanner can detect such wall thickening in the setting of HFH, a model system for the study of atherosclerosis.
In summary, we have shown that patients with HFH have a thicker-walled aortic root, supravalvular aortic narrowing, and smaller aortas than normal age-matched control subjects and that conventional fat-suppressed ECG-gated spin-echo MRI was optimal for visualizing wall thickening due to atherosclerotic plaque. Moreover, the significant reduction in the OD in these patients implies a direct or indirect role for the LDL receptor in aortic growth and development.
Selected Abbreviations and Acronyms
|BSA||=||body surface area|
|FOV||=||field of view|
|HFH||=||homozygous familial hypercholesterolemia|
|NEX||=||number of excitations|
We thank Andrew Dwyer, MD, for critical review of the manuscript and Kim Gallagher for assistance with data entry and manuscript preparation.
Presented at the Fourth Scientific Meeting of the International Society for Magnetic Resonance in Medicine, New York, NY, April 27–May 3, 1996, and NIH Clinical Center Grand Rounds, Bethesda, Md, April 9, 1997.
- Received October 13, 1997.
- Revision received March 23, 1998.
- Accepted April 5, 1998.
- Copyright © 1998 by American Heart Association
Goldstein JL, Hobbs HH, Brown MS. Familial hypercholesterolemia. In: Scriver CR, Beaudet AL, Sly WS, eds. The Metabolic and Molecular Bases of Inherited Disease. New York, NY: McGraw-Hill; 1995:1981–2030.
Ribeiro P, Shapiro LM, Gonzalez A, Thompson GR, Oakley CM. Cross sectional echocardiographic assessment of the aortic root and coronary ostial stenosis in familial hypercholesterolaemia. Br Heart J. 1983;50:432–437.
Hoeg JM, Feuerstein IM, Tucker EE. Detection and quantitation of calcific atherosclerosis by ultrafast computed tomography in children and young adults with homozygous familial hypercholesterolemia. Arterioscler Thromb. 1994;14:1066–1074.
Barkow R, Fletcher AJ. Merck Manual. 15th ed. Rahway, NJ: Merck; 1987.
Zar JH. Biostatistical Analysis. 2nd ed. Englewood Cliffs, NJ: Prentice Hall; 1984.
Allen JM, Thompson GR, Myant NB, Steiner R, Oakley CM. Cardiovascular complications of homozygous familial hypercholesterolaemia. Br Heart J. 1980;44:361–368.
Beppu S, Minura Y, Sakakibara H, Nagata S, Park YD, Nambu S, Yamamoto A. Supravalvular aortic stenosis and coronary ostial stenosis in familial hypercholesterolemia: two-dimensional echocardiographic assessment. Circulation. 1983;67:878–884.
Pak YK, Kanuck MP, Berrios D, Briggs MR, Cooper AD, Ellsworth JL. Activation of LDL receptor gene expression in HepG2 cells by hepatocyte growth factor. J Lipid Res. 1996;37:985–998.
Rechtoris C, Mazzone T. Isoform-specific induction of the low-density lipoprotein receptor gene by platelet-derived growth factor. Am J Physiol. 1995;268:C1033–C1039.
Merickel MB, Berr S, Spetz K, Jackson TR, Snell J, Gillies P, Shimshick E, Hainer J, Brookeman JR, Ayers CR. Noninvasive quantitative evaluation of atherosclerosis using MRI and image analysis. Arterioscler Thromb. 1993;13:1180–1186.
Toussaint JF, LaMuraglia GM, Southern JF, Fuster V, Kantor HL. Magnetic resonance images lipid, fibrous, calcified, hemorrhagic, and thrombotic components of human atherosclerosis in vivo. Circulation. 1996;94:932–938.