Author + information
- Received December 20, 1999
- Revision received November 3, 2000
- Accepted December 12, 2000
- Published online March 15, 2001.
- Gérard Helft, MD, PhD∗,†,
- Stephen G Worthley, MBBS, FRACP∗,†,
- Valentin Fuster, MD, PhD, FACC†,
- Azfar G Zaman, MD, MRCP∗,†,
- Clyde Schechter, MD§,
- Julio I Osende, MD∗,†,
- Oswaldo J Rodriguez, MD∗,†,
- Zahi A Fayad, PhD†,
- John T Fallon, MD, PhD†,‡ and
- Juan J Badimon, PhD, FACC∗,†,* ()
- ↵*Reprint requests and correspondence:
Dr. Juan J. Badimon, Cardiovascular Biology Research Laboratory, Zena and Michael A. Wiener Cardiovascular Institute, Mount Sinai School of Medicine, New York, New York 10029-6574
We sought to demonstrate the ability that noninvasive in vivo magnetic resonance imaging (MRI) has to quantify the different components within atherosclerotic plaque.
Atherosclerotic plaque composition plays a critical role in both lesion stability and subsequent thrombogenicity. Noninvasive MRI is a promising tool for the characterization of plaque composition.
Thoracic and abdominal aortic atherosclerotic lesions were induced in rabbits (n = 5). Nine months later, MRI was performed in a 1.5T system. Fast spin-echo sequences (proton density-weighted and T2-weighted [T2W] images) were obtained (in-plane resolution: 350 × 350 microns, slice thickness: 3 mm). Magnetic resonance images were correlated with matched histopathological sections (n = 108).
A significant correlation (p < 0.001) was observed for mean wall thickness and vessel wall area between MRI and histopathology (r = 0.87 and r = 0.85, respectively). The correlation was also present on subanalysis of the thoracic and upper part of the abdominal aorta, susceptible to respiratory motion artifacts. There was a significant correlation for plaque composition (p < 0.05) between MRI and histopathology for the analysis of lipidic (low signal on T2W, r = 0.81) and fibrous (high signal on T2W, r = 0.86) areas with Oil Red O staining. T2-weighted images showed greater contrast than proton density-weighted between these different components of the plaques as assessed by signal intensity ratio analysis with the mean difference in signal ratios of 0.47 (S.E. 0.012, adjusted for clustering of observations within lesions) being significantly different from 0 (t1= 39.1, p = 0.016).
In vivo noninvasive high resolution MRI accurately quantifies fibrotic and lipidic components of atherosclerosis in this model. This may permit the serial analysis of therapeutic strategies on atherosclerotic plaque stabilization.
Atherosclerotic disease is one of the leading causes of morbidity and mortality in Western societies (1–3). Atherosclerotic plaque composition, rather than stenotic severity, has emerged as the most important determinant for the thrombus-mediated acute coronary syndromes. The so-called “vulnerable” atherosclerotic plaques, characterized by a large lipid core surrounded by a thin fibrous seem to play a critical role in the onset of acute coronary syndromes (4,5). Therapeutic approaches for the prevention of acute coronary syndromes appear to act by stabilizing the atheromatous plaque through modification of lesion composition (6,7).
The possibility of an imaging modality able to quantify plaque composition would allow the stratification of patients at high risk for plaque rupture. Different imaging modalities including magnetic resonance imaging (MRI), angioscopy, vascular ultrasonography and infrared imaging analysis can differentiate between normal and atheromatous arteries (8–12). Magnetic resonance (MR) techniques appear to have the ability to noninvasively identify and discriminate the components of complex atherosclerotic lesions both in ex vivo and in vivo settings (8,9,13–18)but have not been shown to allow quantification of the plaque components. Plaque characterization, in particular lipidic versus fibrotic components, is critical in predicting which lesions are at risk for disruption; therefore, the assessment of plaque composition could have significant clinical and therapeutic implications.
We report the feasibility of in vivo noninvasive high resolution MRI to characterize and quantify lipidic and fibrotic components in aortic lesions in a rabbit model of atherosclerosis. We also demonstrate the utility of T2-weighted (T2W) images in the differentiation of these components within the plaque. This method provides the basis for the continued development and investigation of the use of MRI to noninvasively assess changes in the composition of atherosclerotic lesions.
Animal model of experimental atherosclerosis
The animal model selected for this study was the New Zealand white rabbit (n = 5, weight 3.0 to 3.5 kg). Atherosclerotic aortic lesions with fibrotic and lipidic components were induced by a combination of atherogenic diet (0.2% cholesterol) for nine months and repeat balloon denudation, as previously described (18). All experiments were approved by the Institutional Animal Management Program.
Nine months after the initiation of the atherogenic diet, rabbits were anesthetized and placed supine in a 1.5 Tesla clinical MRI system (Signa, General Electric) using a conventional phased-array volume coil (18). Gradient-echo coronal images were used to localize the thoracic and abdominal aorta. Thereafter, sequential axial images (3-mm thickness) of the aorta from the arch to the iliac bifurcation were obtained using a fast spin-echo sequence (total imaging time: 70 min) with an in-plane resolution of 350 × 350 microns (proton density-weighted [PDW]: TR/TE: 2,300/17 ms; T2W: TR/TE: 2,300/60 ms, field of view = 9 × 9 cm, matrix 256 × 256, echo train length = 8, signal averages = 4).
Rabbits were euthanized within 48 h of MR as previously described (18). After the removal and parafolmaldehyde (4% in PBS) fixation, serial sections of the aorta were cut at 3-mm intervals matching corresponding MR images. The total number of sections analyzed was 108, n = 53 for the thoracic and supra-renal aortic sections, 20 to 25 sections per rabbit. Coregistration was performed carefully by utilizing the position of the aortic arch, renal arteries and iliac bifurcation. The selected aortic specimens were paraffin-embedded, and 5 μ thick sections were cut and stained with combined Masson’s trichrome elastin stain. A subset of abdominal aortic specimens were kept at −80°C for specific lipid staining with Oil Red O (n = 7).
Image and data analysis
The MR images were transferred to a Macintosh computer for further analysis. The histopathological sections were digitized to the same computer from a camera (Sony, 3CCD Video Camera) attached to a Zeiss Axioskop light microscope. The MR images were matched with corresponding histopathological sections for the aortic specimens (n = 108).
Cross-sectional areas of the lumen and outer boundary of each aortic section were determined for both MR images and histopathology by manual tracing with ImageProPlus (Media Cybernetics) as previously reported (18). From these measurements, mean wall thickness (MWT), vessel wall area (VWA) and lumen area were calculated. In those MR images corresponding to the subset of atherosclerotic plaques (n = 7) prepared by frozen sectioning and staining with Oil Red O, lipidic and fibrotic areas were measured by manual tracing as described above. Mean wall thickness and VWA were analyzed with combined Masson elastin staining; lipidic and fibrotic areas were analyzed with Oil Red O staining. The analyses were blindly performed by two independent investigators.
Aortic sections in which fibrotic and lipidic components could be separately identified on the MR image (n = 23) were further analyzed in order to assess the contrast ratios between fibrotic and lipidic components for PDW and T2W images. For each MR image of the complex plaques, signal intensity was measured for PDW and T2W images at two separate points within both the fibrotic (defined as high-signal regions) and lipidic (defined as low-signal regions) areas as previously reported (13,18). This distinction by MR was confirmed by combined Masson elastin staining.
Comparisons of histological and MRI-derived vessel-wall measurements were performed using Stata v. 6.0 (Stata Corporation, College Station, Texas). Linear regressions were carried out using the MRI-derived measurement as the dependent variable and the corresponding histological measurement as the independent variable. Design effects to account for multiple sections derived from each rabbit were calculated and applied using the Stata’s svyreg procedure.
The ratio of PDW and T2W signal-intensities were calculated in the fibrous and lipidic lesion areas. Comparison of the PDW fibrous/lipid signal intensity ratio to the T2 signal intensity ratio was analyzed by linear regression, with the ratio as the dependent variable and the weighting method (PDW or T2) as the independent variable. Design effects to account for multiple observations derived from each lesion were again calculated and applied using the Stata’s svyreg procedure.
To define intraobserver and interobserver variability for the quantification of lipidic and fibrotic components, the MR images of aortic atherosclerotic plaques used for comparison with Oil Red O staining (n = 7) were reanalyzed and the intraclass correlation coefficients determined.
Atherosclerotic characterization by histopathology
The combination of the atherogenic diet for nine months and repeat balloon denudation of the whole aorta induced the formation of raised lesions characterized by an increase in lipid and fibrotic components. However, unlike humans, no calcium or intraplaque thrombus is associated with this model. The staining clearly allowed the discrimination of lipidic and fibrotic areas. The lipidic areas were mainly located in the deeper part of the wall, while the fibrotic areas showed a more luminal distribution (Fig. 1). Histopathology sections were compared with MRI for MWT and VWA (n = 108) and atherosclerotic component quantification (n = 7), and MR images with both fibrotic and lipidic components (n = 23) underwent signal intensity analysis.
The PDW and T2W axial images showed marked wall thickening in the aorta, from the aortic arch to the iliac bifurcation. In these in vivo images, the collagen-rich areas appeared brighter than the foam-cell rich areas seen darker. There was a good correlation between the images obtained in vivo and the histopathological sections of the aorta (Fig. 1). Lesion coregistration was obtained by utilizing the position of the aortic arch, renal arteries and iliac bifurcation (Fig. 2).
Quantification of plaque composition
The fast spin-echo images showed different signal intensity areas identifying fibrous and lipidic plaque components. A relatively dark area (low-signal intensity both on T2W and PDW) and a bright area adjacent to the lumen (high-signal intensity both on T2W and PDW) were clearly distinguished (Fig. 3). There was a significant correlation (p < 0.05) between T2W images and histopathology for analysis of lipidic and fibrous areas (r = 0.81, y[MR] = 1.25 + 1.02x[histo] and r = 0.86, y[MR] = 0.53 + 1.35x[histo], respectively). We did note, however, a consistent overestimation of all arterial wall parameters by MRI when compared with histopathology.
Signal intensity analysis
The mean fibrous/lipid signal ratio for PDW signals was 1.36 and that of T2 signals 1.83. The mean difference in signal ratios is 0.47 (S.E. 0.012, adjusted for clustering of observations within lesions) and is significantly different from 0, t1= 39.1, p = 0.016, indicating that T2W discriminate better between fibrous and fatty components of lesions.
Atherosclerotic wall measurements
When all aortic sections are considered, there was a strong association between the MRI- and the histology-derived measurements of MWT ([MRI] = 0.17 + 0.83 × MWT[hist], R2= 0.75, F1,4= 281.43, p = 0.0001). Similarly, a strong correlation between MRI- and the histology-derived measurement of VWA ([MRI] = 0.22 + 1.49 × VWA[hist], R2= 0.71, F1,4= 64.99, p = 0.0013). The correlation was still observed considering only the thoracic and upper part of the abdominal aorta (Fig. 4)susceptible to respiratory motion artifacts: MWT (MRI) = 0.28 + 0.65 × MWT(hist) and MWT(VWA) = 2.02 + 1.37× VWA(MRI).
Intra- and interobserver variability assessment by intraclass correlation for MRI of fibrotic and lipidic components showed good reproducibility, with the intraclass correlation coefficient ranging from 0.91 to 0.98.
Our study demonstrated that in vivo MRI can reliably and noninvasively detect and quantify fibrous and lipid components of aortic atherosclerotic lesions in the rabbit model. We have shown the superiority of T2W over PDW imaging in the discrimination of the different components within the plaques. Despite respiratory and cardiac motions, thoracic and abdominal aortic lesions were well characterized in this rabbit model.
Studies of coronary disease have shown that its progression is unpredictable by angiography (19,20). The sites at which the thrombotic lesions occurred are usually of mild-to-moderate stenotic severity (21). The combination of a large core and a thin cap is the major determinant of plaque vulnerability (5,22). A noninvasive imaging technique capable of assessing not only the measurements of the wall thickness and VWA but also measurements of different components of the atherosclerotic lesions is crucial for the risk stratification of patients based on plaque vulnerability. Currently, we have compared MR images obtained with a clinical magnet and a conventional coil to histopathological observations. To date, there is no imaging technique that has demonstrated its ability to measure and quantify the different components of atherosclerotic plaques.
Plaque characteristics with MRI
We have been able to accurately assess plaque size and discriminate and quantify the fibrotic and lipid components of plaques. The lipid areas, having a low signal on T2W and PDW images, are very distinctive from the fibrous areas, which have a high signal on T2W and PDW images (13,23). Proton density-weighted images with an echo time of 17 ms have some T2 character. This could explain the low signal of the lipid areas with PDW imaging. We are reporting a significant correlation between MR imaging and Oil Red O histopathology for analysis of lipid (r = 0.81) and fibrous (r = 0.86) areas. The fibrous cap as identified by MRI appears thicker than that by histopathology. The fibrous cap is thin (around 0.25 to 0.5 mm) and with an in-plane resolution of 0.35 mm; clearly there is a tendency towards overestimation with this MR technique, a concept related to “partial volume averaging.” Also, the histopathological specimens are an imperfect representation of the in vivo tissue, with vessel shrinkage due to desiccation during specimen preparation (15,16,18). The lipid areas are larger, as per histopathology, and, thus, less subject to overestimation by MRI with the resolution described. Importantly, despite these potential limitations, the good correlation observed between the two techniques strongly validates the potential value of MRI for serially documenting in vivo changes over time.
In this model, T2W images had greater contrast between the lipid and fibrotic areas within the plaque than the PDW images. Previous reports concluded that PDW images are the most useful for demonstrating atherosclerotic plaque components in a rabbit model (23). Our data show that T2W images have the highest contrast between these components of the plaques. For comparability, we used the same echo time (17 ms) for our PDW imaging as other studies (23).
The analysis of MWT and VWA demonstrated a strong correlation between MRI and histopathology observation. Previous authors have only focused on identification of plaques in the abdominal aorta of rabbits due to cardiac and respiratory motions (13,24). We found that the thoracic aorta, which is adherent to the paraspinal structures, was relatively spared from these artifacts. Mean wall thickness and VWA showed a significant correlation between MRI and histopathology for the thoracic and supra-renal aortic sections, a region of the aorta susceptible to respiratory artifacts from the diaphragm.
Our study was conducted with a conventional coil and on a whole-body 1.5 Tesla clinical scanner. This indicates the applicability of this technique to the study of human atherosclerosis. The use of two-dimensional imaging techniques, such as that used in this study, limits the “through-plane” or “z-axis” resolution. Therefore, future efforts to improve resolution in all three axes using overlapping two-dimensional MR slices or especially using a three-dimensional imaging sequence warrant further study. In the future, the measurements of the different components of the plaque could provide a new prognostic tool useful in the management of atherosclerosis.
In conclusion, our data indicate that MRI has the potential to noninvasively quantify lipid and fibrotic areas in vivo in atherosclerotic plaques using current clinical MR techniques. This method provides the basis for the continued development and investigation of the use of MRI to noninvasively assess changes in the composition of atherosclerotic lesions.
☆ Supported by grants from the French Federation of Cardiology (G.H.), the National Heart Foundation of Australia (SA Branch) (S.G.W.), the Spanish Society of Cardiology (J.I.O.) and the National Institutes of Health (NIH P50 HL54469) (V.F., J.T.F. and J.J.B.).
- magnetic resonance
- magnetic resonance imaging
- mean wall thickness
- proton density-weighted
- vessel wall area
- Received December 20, 1999.
- Revision received November 3, 2000.
- Accepted December 12, 2000.
- American College of Cardiology
- Davies M.J.
- Falk E.,
- Shah P.K.,
- Fuster V.
- Libby P.
- Toussaint J.F.,
- LaMuraglia G.M.,
- Southern J.F.,
- Fuster V.,
- Kantor H.L.
- Brezinski M.E.,
- Tearney G.J.,
- Weissman N.J.,
- et al.
- Fayad Z.A.,
- Fallon J.T.,
- Shinnar M.,
- et al.
- Worthley S.G.,
- Helft G.,
- Fuster V.,
- et al.
- McConnell M.V.,
- Aikawa M.,
- Maier S.E.,
- Ganz P.,
- Libby P.,
- Lee R.T.
- Worthley S.G.,
- Helft G.,
- Fuster V.,
- et al.
- Lichtlen P.R.,
- Nikutta P.,
- Jost S.,
- Deckers J.,
- Wiese B.,
- Rafflenbeul W.
- Mann J.M.,
- Davies M.J.