In Vitro and In Situ Magnetic Resonance Imaging Signal Features of Atherosclerotic Plaque-Associated Lipids
Abstract The goal of this study was to evaluate magnetic resonance imaging (MRI) signal features of the different types of lipids found in human atherosclerotic plaques. A 1.5-T SIGNA scanner was used to acquire T1-, T2-, and proton density– weighted data at four different temperatures for individual lipids and lipid mixtures designed to replicate the proportions of lipids found in plaques. Individual lipids and lipid mixtures were scanned both in a test tube and after implantation in the media of normal porcine aortas. Each of the three broad classes of lipids (triglycerides, unesterified and esterified cholesterol, and phospholipids) had different and distinct MR signal patterns, which allowed discrimination of these classes of lipids in vitro. Further, lipid implantation studies demonstrated that these distinct MR signal patterns could be used to readily distinguish each lipid type from surrounding porcine aortic media. MR signals from lipid mixtures demonstrated marked regional heterogeneity, similar to the heterogeneous lipid distribution characteristic of human atherosclerotic plaques. In summary, MR signals from lipid mixtures that mimic plaque lipid proportions can be detected at body temperature, especially in those mixtures with an increased percentage of cholesteryl esters. These studies raise the possibility that with further advances in technology, MRI may become a useful tool for determining the lipid content and composition of human atherosclerotic plaques in vivo.
- Received December 29, 1995.
- Accepted October 25, 1996.
Atherosclerotic cardiovascular disease is one of the leading causes of morbidity and mortality in Western societies. Recent studies have demonstrated that most clinical cardiovascular events result from plaque ulceration, intraplaque hemorrhage, and/or plaque rupture, all of which can result in acute narrowing or occlusion of the arterial lumen. A number of recent studies have also linked the presence of a lipid-rich atherosclerotic plaque core to plaque instability and rupture.1 2 3 4 5 6 Davies et al3 suggested that in aorta, ulceration and thrombosis were characteristic of plaques with a high proportion of their volume occupied by extracellular lipid. Falk4 and Leen et al5 theorized that the formation of a lipid core and the weakening of the overlying fibrous cap may predispose the plaque to disruption. Stary7 and Guyton et al8 found that the growth of the lipid core may be associated with progressive tissue destruction and necrosis within the plaque. Davies et al,3 Richardson et al,6 and Leen et al5 have suggested that the presence of a lipid core increases the mechanical stress on the fibrous cap, thereby increasing its likelihood to rupture. At least two lines of evidence suggest that the plaque lipid core also may increase the likelihood of plaque rupture by serving as a nidus for the recruitment of protease-producing macrophages. Lendon et al9 found that, compared with the fibrous caps of intact plaques, the fibrous caps of ruptured plaques had higher macrophage densities. In addition, Nikkari et al10 have demonstrated recently that macrophages expressing interstitial collagenase, an enzyme that degrades the major structural collagen types found in plaques, are particularly prominent at the border of the plaque lipid core and correlate with the amount of intraplaque hemorrhage. Recent studies in patients with hyperlipidemia and coronary disease have suggested that reduction in plasma LDL cholesterol, known experimentally to deplete lipid from plaque core and macrophage pools, effectively stabilizes these high-risk plaques.11
The lipid core of atherosclerotic plaques consists of cholesteryl esters, cholesterol monohydrate, phospholipid, and a small amount of triglyceride and has been found to undergo both chemical and physical changes in composition as atherosclerotic lesions progress.12 13 14 15 Small and Shipley14 found that the percent dry weight of lipids increases as lesions progress from normal intima (1.2%) to fatty streaks (25%) and then to atheromatous (gruel) plaques (60%). Concurrently, the cholesterol monohydrate component of the plaque increases and precipitates as crystals, while the cholesteryl ester component, found as droplets, decreases.13 15 If the presence and distribution of these lipids could be monitored with a noninvasive imaging modality such as MRI, then it might be possible to identify patients with plaques at high risk for various manifestations of plaque instability (ulceration, intraplaque hemorrhage, and fibrous cap rupture) that result in clinical cardiovascular events. Compared with the conventional imaging modalities used to study atherosclerotic plaques (including contrast angiography and ultrasound), MRI provides better soft-tissue contrast. Its potential to determine both the lumen size and the constituents of atherosclerotic plaques has been demonstrated previously.16 17 18 19 20 21
However, several studies have reached contradictory conclusions about the utility of MRI in identifying plaque lipid components. Shahrokh et al22 found that the plaque lipid components contribute to the T2 signals of plaques when they are in their liquid phase. These investigators also suggested that since lipids are a small fraction of the total plaque compared with water and that the majority of the lipids present in plaques are not liquid, the ability to monitor the lipid component of plaques by MRI may be limited. Martin et al16 used T2W imaging to study plaques and concluded that plaque lipids are invisible to imaging. However, other studies have indicated that in contrast to solid-phase lipids, atheromatous lipids in a liquid crystalline phase have longer T2 values and are visible on MR images.19 23 Thus, the inability of earlier investigators to image plaque lipids may be due not to a lack of utility of MRI but rather to an incomplete appreciation of the marked heterogeneity in the types and phases of lipids present in human atherosclerotic plaques.
Because of the clinical potential for lipid imaging, it is important to develop our understanding of the similarities and differences in MR signal patterns for specific plaque-associated lipids and of how these plaque lipid MR signal patterns may differ from those of arterial regions without lipid accumulation as well as those of perivascular fat. The purpose of this study was to evaluate the MRI signal patterns of individual lipids implanted in the media of normal porcine aorta in volumes approximating the core lipid region of human atherosclerotic plaques. The most prominent lipids found in human plaques were studied at four temperatures. MRI signal differences between the anhydrous and hydrated lipids also were studied. In addition, the signal patterns of mixtures of lipids that mimic the known concentrations of lipids found in plaques were studied at the same four temperatures.15 Another purpose of this study was to determine whether individual types of lipids could be distinguished by determining the contrast patterns for each in T1W, T2W, and PDW MR images.
Fresh 15- to 20-cm segments of porcine abdominal aortas without macroscopic evidence of plaque were obtained from healthy 2-year-old pigs (Kapowsin Meats, Graham, Wash) and stored at −10°C until used for study. For each experiment, the aorta was allowed to thaw, then cut axially into 5- to 7-cm sections, dissected sagittally, and trimmed of most perivascular fat. A custom-built plastic platform was made to support the sections for imaging (Fig 1⇓).24 The plastic used throughout the 5×10-cm platform had minimal MR signal. Clamps, made of two long plastic strips bound on either end by plastic screws, were used to bind the axially spread aorta on the left and right side, displaying the endothelial surface in the ventral direction (Fig 1⇓). Lipid injections then were performed.
Preparation and Injection of Individual Lipids
A series of representative lipids of atherosclerotic plaques were injected into the abdominal aortas. These lipids and their melting points and grade are listed in Table 1⇓. The lipids, with the exception of triolein, were obtained commercially in purified form as a crystalline powder (Sigma Chemical Company). Triolein was obtained in liquid phase from the same company. The lipids were injected into the aortas in three ways: (1) as lipid rods through a needle, using a technique described by Brown and Fry,25 (2) as solid crystals through slits in the aortas, and (3) as liquid through a needle.
The lipid rods of individual lipid types were fabricated by using blunt-ended inserters of 22-gauge spinal needles to pack the powdered lipids into normal 18-gauge needles. The lipid rods varied from 3 to 5 mm in length and from 1 to 2 mm in diameter. Under light microscopy, the crystals within lipid rods were observed to be randomly oriented.
The lipid rods were injected into the luminal side of the platform-bound aortas in the following manner: The lipid-containing 18-gauge needle was inserted in a cephalad direction as superficially as possible in a plane of dissection parallel to the endothelium. Once the needle tip was adequately advanced under the intima (5 to 12 mm), a 22-gauge spinal needle was used to push the lipid rod to the needle tip. While the 22-gauge spinal needle was held stationary, the 18-gauge needle was withdrawn. The 22-gauge spinal needle was then removed, leaving the lipid rod in place. There was a 1- to 4-mm space of tissue between the lipid rods and the needle injection site. The lipid rods were placed in rows inside the aorta.
In the case of liquid lipids, the lipids were drawn into 1.0-mL insulin syringes. The needles were inserted in the same manner as the lipid rods, advanced 4 to 5 mm, and the lipid was injected. While the needle was held in place, a suture was sewn around the needle and drawn tight. As the needle was retracted, the suture continued to be drawn tight, to close the needle hole and minimize leakage.
Solid crystals were inserted by making slits with a scalpel as superficially as possible in a plane of dissection just underneath and parallel to the aortic luminal endothelium. The slit was opened using clamps, and the mixtures of stimulated plaque lipids were placed as far down into the slits as possible. The slits then were closed using 6-0 monofilament sutures.
Preparation of Lipids for the Comparison Between Hydrated and Anhydrous States
Lipids in atherosclerotic tissue are in contact with an aqueous milieu, ie, are hydrated, while crystalline lipids obtained commercially are anhydrous. To obtain lipids exposed to water and therefore more likely to reflect the status of lipids in the aqueous environment of plaques, anhydrous lipids were combined, in microfuge tubes, with equal amounts of distilled H2O (50%:50% wt/vol). Each lipid/H2O combination then was heated with a heating block until the lipid was observed to melt or to a temperature of 110°C, whichever came first. At that point, the lipid/H2O combination was mixed by vigorous vortexing.
Preparation of Lipids Mixture
Samples (50 mg) of lipid mixtures that simulated the proportions of lipids found in plaques (atheroma) with high, low, and normal cholesteryl ester content were prepared using compositions determined for these different plaque types by Small.15 Increases in the proportion of total cholesteryl esters were coupled with corresponding decreases in the total proportion of unesterified cholesterol, while the relative proportions of sphingomyelin, triolein, and l-α-lecithin were kept constant (Table 2⇓). Due to the concerns of accuracy of weighing small amounts of compounds, the number of cholesteryl ester types was limited to three (cholesteryl oleate, linoleate, and stearate). These compounds represent the most abundant cholesteryl esters in plaques and have a range of melting points representative of the other, less prevalent, types. To determine how much of each of the three major cholesteryl ester types to add, the melting points of all cholesteryl esters listed by Small were determined and the amounts of these minor cholesteryl esters were substituted for whichever of the three major cholesteryl esters had the closest melting point.15 Lysolecithin was represented by l-α-lecithin, and triglycerides were represented by triolein, which is liquid at room temperature. The same hydration and heating process used for individual lipids was used for the lipid mixtures, which were placed in microfuge tubes.
The lipid-injected aorta or lipid-filled tubes were placed into a polyethylene specimen container for MRI. This container was built to accommodate the platform, a water solution for temperature control, and a custom-built integrated RF surface coil (Fig 1⇑).24 A tube passing through the container was used to control the sample temperature by connecting to a heat bath. The experimental temperatures of the container were calibrated individually.
The MR images were taken from a 1.5-T whole-body SIGNA imager, version 5.4 (GE Medical Systems). A two-dimensional spin echo sequence was used to acquire T1W, T2W, and PDW images across regions with lipid injections. The imaging parameters (TR [ms]/TE [ms]/field-of-view [cm]/matrix size/slice thickness [mm]/excitation) for T1W were 600/20/8/512×256/1.5/2; for T2W, 2500/80/8/512×256/1.5/2; and for PDW, 2500/20/8/512×256/1.5/2. For each lipid-injected aorta, T1W, T2W, and PDW images were acquired at four different temperatures (37°C, 40°C, 45°C, and 50°C). Sample images are presented in Fig 2⇓.
For injected lipids, signal intensities were measured on the MR images from sites of injection, from uninjected aorta, and from the platform using the Independent Console (GE Medical Systems). Each image was magnified by a factor of two and the signal intensities were measured from all lipid areas by using an ROI of 0.122 mm2. For each lipid site, multiple intensity readings were taken, with the lowest reading being chosen to minimize the partial volume effects of the surrounding tissue. To determine the background, signal intensities of a 10- to 12-mm2 area were taken from the platform.
The CNR ratio was determined by the formula CNRi=(Si−So)/SD, where Si is the signal intensity of lipid/simulated plaque site i, So is the signal intensity of the platform, and SD is the noise equal to the SD of the background air. For each single lipid or simulated plaque, a mean CNR and SD were calculated on the basis of measurements from a minimum of five sites.
For lipids contained in the microfuge tubes, the measurements conducted were the same as for the injected lipid model, with the exception that the size of ROIs varied according to the amount of lipid presented. Due to the positioning of different vials, a surface coil correction program was used to remove signal variations due to coil sensitivity before measurements.26 A mean CNR and SD were calculated on the basis of measurements from a minimum of five sites.
Fig 3⇓ shows the CNR changes over different temperatures for single lipids in T1W, T2W, and PDW images. Triolein, which is in liquid phase at body temperature, showed the highest signal intensity and changed little with increasing temperature. The artery wall, into which the lipids were injected, had CNR values that were slightly less than triolein in PDW and T2W images and half those of triolein in T1W images. Signal intensities for cholesteryl stearate, cholesterol, and l-α-lecithin varied little over the temperature range studied. Cholesteryl linoleate, with a melting point of 42°C, showed a dramatic increase in signal intensity in all three contrast weightings after the experimental temperature passed its melting point. Similarly, cholesteryl oleate (melting point of 46°C) showed a similar increase in signal strength after it passed its melting point. Sphingomyelin had relatively large T1W and PDW signal strengths but a weak T2W signal strength. The CNR of the artery wall changed little in T1W and PDW images and decreased in T2W images with increasing temperature.
Signal Differences Between Lipids in Anhydrous and Hydrated State
Compared with their anhydrous counterparts, significant signal increase was observed with hydration for sphingomyelin and l-α-lecithin in T1W, T2W, and PDW images (Fig 4⇓). In addition, the CNR features of anhydrous sphingomyelin were different from those shown in Fig 3⇑, likely indicating that the sphingomyelin injected into aorta in the anhydrous state had become partially hydrated by the time of imaging. Once hydrated, the CNR values of sphingomyelin were strong in all three contrast weightings, especially in T2W imaging. The CNR values of hydrated l-α-lecithin were significantly larger than those of anhydrous state but less than the CNRs of sphingomyelin. No significant signal changes were observed with hydration for unesterified cholesterol and/or any of the three cholesteryl esters tested (data not shown). Triolein was not tested, since it already was in liquid phase at the temperatures studied.
Overall, at body temperature, among the individual lipids, the highest MR signal intensities were obtained from the triglyceride (triolein) and from the phospholipids (sphingomyelin and l-α-lecithin). The signal intensities of cholesterol and cholesteryl esters were quite low. Because the signal intensities of normal aorta were intermediate between the high triglyceride and phospholipid signals and the low signal intensity of cholesterol and cholesteryl esters, all four of these groups could be distinguished from normal aortic tissue (P<.02). T1W imaging will be the choice to differentiate aortas and single lipids, as evidenced by the contrast differences between aortas, phospholipid, triglyceride, and cholesteryl esters.
Signal intensities of lipid mixtures that simulate the proportions of different lipids found in plaques showed marked regional heterogeneity (Fig 5⇓). This heterogeneity was more prominent in the two lipid mixtures with the lower cholesteryl ester (and therefore the higher unesterified cholesterol) contents. Fig 5⇓ shows a group of images of the three contrast weightings taken from three sequential (equal distanced) locations of microfuge tubes containing the three lipid mixtures. A microfuge tube filled with H2O was also included as reference. In these images, high signal intensity regions were seen in all three types of mixtures interleaved with regions of low signal intensity. These results indicate that the different lipid types were not homogeneously mixed in these lipid mixtures designed to simulate plaque lipid proportions. Despite these signal variations, the overall signal intensities were higher in both T1W and PDW images for lipid mixtures with high cholesteryl ester content and lower for lipid mixtures with lower cholesteryl ester content.
This study used lipid CNR values to express the MR signal patterns of cholesterol, cholesteryl esters (cholesteryl linoleate, stearate, and oleate), triglyceride (triolein), and phospholipid (l-α-lecithin and sphingomyelin) at different temperatures to establish a realistic model for preclinical testing. The contrast comparison was relative to the plastic platform, which had a signal intensity similar to air. Thus, a CNR value of zero corresponds to very low signal intensity.
The images taken from normal porcine aortas with implants of individual lipids demonstrated that each of the lipid implants could be distinguished from normal aortic media in each of the three contrast weightings tested at the spatial resolution used. This spatial resolution (0.16 mm×0.31 mm) is currently used in in vivo human carotid artery imaging.17 Thus, our findings may eventually have direct clinical applications. For example, an important next step will be to assess the signal feature differences between lipid cores of atherosclerotic plaques and the fibrous caps that separate the core from flowing blood. In addition, other practical issues of in vivo imaging will need to be considered, including the effects of vessel motion due to pulsatile blood flow, the structural complexity of normal and atherosclerotic human artery walls, and improving spatial resolution. These problems have been addressed to some extent for carotid arteries,17 but attempts to extend these techniques to coronary arteries will provide additional challenges, such as increased vessel motion due to cardiac contraction and a need for higher spatial resolution due to smaller vessel size.
The CNR values for individual lipids shown in Fig 3⇑ indicated that, with the exception of sphingomyelin, the individual lipids did not contribute to MR signals until the imaging temperature reached or exceeded their melting point. This is evidenced in CNR changes of cholesteryl linoleate and oleate, both of which have melting points between 40°C and 50°C. This result is expected, because once in the liquid phase, lipid molecule motion increases to a rate near the MR Larmor frequency. This results in faster energy exchange among molecules, which translates into a shorter T1 relaxation time and higher MR signals.27 The signal contribution of liquefied cholesteryl linoleate and oleate may be higher than the results shown in Fig 3⇑. The slightly lower signal could be partially due to the fact that only the lowest reading from the lipid site was used for contrast-to-noise calculation.
Triolein, a triglyceride in liquid phase at body temperature, had strong signals compared with both porcine aortas and other type of lipids in all three contrast weightings. Further, there was little change in signal intensity over different temperatures. These signal features are similar to the adipose tissue signal observed in MR images. Aortas, with media as the main constituent, contain smooth muscle cells, elastin, and collagen. Their MR signals tend to be similar to those of muscular tissues.16 Thus, in T1W images, triolein had stronger (and hyperintensive) signals than aortas, and in T2W and PDW images, the signals of triolein and aortas are comparable.
Signal Differences Between Lipids of Anhydrous and Hydrated State
The physical state of lipid can have significant impact on MR signal intensities, as demonstrated in this study. Since the hydrated state is the normal state of lipids in human atherosclerotic plaques, the signal features obtained from hydrated lipids are more relevant to gauging the results that might be expected from in vivo imaging. Among all the lipids tested, including cholesterol, cholesteryl esters, and phospholipids, significant increases of signal intensities attributed to hydration were observed in the two phospholipids (sphingomyelin and l-α-lecithin).
In contrast to other lipids, which are almost entirely hydrophobic, both sphingomyelin and l-α-lecithin are amphiphilic, possessing both hydrophilic polar head groups and hydrophobic hydrocarbon tails. The polar head groups of these phospholipids allow them to combine with water, causing them to swell into liquid crystalline or “gel” structures when mixed with water.28 It is likely that this mixture between water and phospholipid molecules produces a proton-interactive environment that generates MR signal features which differ from those of either pure water or lipids, such as triglycerides, which are in the liquid phase at body temperature. The observation that there were no significant changes in the signal intensities of these hydrated phospholipids over the range of temperatures tested suggests that the structure of these gels did not change substantially under the conditions of this experiment.
The stronger signal of sphingomyelin compared with l-α-lecithin may be due to either or a combination of two factors. First, sphingomyelin has greater relative polarity than l-α-lecithin. This property should allow a higher proportion of water molecules to be incorporated into the sphingomyelin compared with the l-α-lecithin gel. Second, because sphingomyelin has two nitrogens per molecule, while l-α-lecithin has only one, sphingomyelin would be expected to have higher nitrogen-related paramagnetic resonance than l-α-lecithin.
Because atherosclerotic plaques contain a mixture of lipids, the MRI contrast features of lipid mixtures simulating plaque lipid proportions have more relevance to potential clinical application of this technology.15 There were three major differences in the MR signals between lipid mixtures and individual lipids: (1) at body temperature, all three lipid mixtures contributed to MR signals, which were more pronounced in T1W and PDW images; (2) signals from the three different lipid mixtures were heterogeneous, more pronounced in mixtures with intermediate and low cholesteryl ester contents; and (3) at body temperature, lipid signals in lipid mixtures with higher cholesteryl ester contents were stronger than those of the other two lipid mixtures in T1W and PDW images and lower in T2W images.
As expected from individual lipid studies, lipid mixtures simulating plaque lipid proportions can generate strong overall MR signals. However, a more important finding is that the signal can be heterogeneous, which strongly suggests that the different types of lipids are not in a uniformly mixed state, even though vigorous vortexing was applied in all cases after samples were heated to 110°C. The regional heterogeneity of the vortexed hydrated lipid mixture could be from differential regional concentrations of the lipid crystals. Thus, a subregion with high signal intensities in T1W images may locally contain more phospholipid, and a subregion with low signal may contain more free cholesterol.
The overall signal intensity of the three simulated plaques measured from a large ROI demonstrated that plaques with higher cholesteryl ester content had stronger signal and that the intensity increased with increasing temperature (data not shown), a strong indication that an increasing amount of cholesteryl esters was changing to liquid phase and thus contributed to the MR signal. This phenomenon of increasing signal intensity with temperature was not apparent in lipid mixtures with intermediate and low cholesteryl ester contents. The most likely explanation for the higher signal intensities of the lipid mixtures with high cholesteryl ester content is that the combination of these constituents lowered the overall melting point. Small28 showed that when cholesteryl linoleate, oleate, and palmitate are mixed at certain combinations, the overall melting point decreases. The same could be true of combinations of cholesteryl linoleate, oleate, and stearate. If a combination of these three cholesteryl esters lowered all of their melting points, then the MR signal intensity would increase even at body temperature.
The signal heterogeneity of mixed lipids raised two important points: (1) It may be possible to further differentiate lipid-rich regions of plaque into subregions of cholesterol, cholesteryl ester, and phospholipids and (2) a single MR contrast weighting may not be enough to identify these regions. For example, the hypointensive region of lipid mixtures with low cholesteryl ester content had similar signal intensity as water in the T1W image (Fig 5a⇑) but was darker than water in the PDW image (Fig 5d⇑). In addition, this region may contain more free cholesterol than its neighboring hyperintensive region. Further studies will be needed to correlate the differences of MR signal from lipid-rich regions of plaques to the differences of lipid contents and to compare the signal features of lipid-rich regions with those of other tissues of atherosclerotic plaques, especially fibrous plaque and intraplaque hemorrhage.
In summary, this study shows that among individual lipids, triglycerides, which constitute 6% of the plaque, have strong signals in all three MR contrast weightings at body temperature. Hydrated phospholipids, typified by sphingomyelin and l-α-lecithin, also have high MR signal intensities. In contrast, cholesterol and cholesteryl esters have very low MR signal at body temperature. Because the MR signals of aortic media are intermediate in intensity, it may be possible to distinguish both high-MR-intensity triglycerides and phospholipids and low-intensity cholesterol and cholesteryl esters in arterial tissue. Further, the marked regional heterogeneity of MR signals seen in lipid mixtures that simulate plaque lipid proportions suggests that it may be possible not only to determine the total lipid content of plaques but also to get some idea of the relative proportions of different lipid types within an individual plaque. It will be important for future studies to determine whether these observations can be confirmed by careful correlation of MR images with histological examination of human atherosclerotic lesions. If this is the case, this technique may have the exciting potential of allowing the detection of lipid-rich as opposed to lipid-poor plaques in vivo to confirm whether these lipid-rich plaques are indeed at higher risk of subsequent rupture and to follow serially the response of such plaques to interventions such as lipid-lowering or antioxidant therapy.
Selected Abbreviations and Acronyms
|ROI||=||region of interest|
|T1W, T2W||=||T1- and T2-weighted|
We gratefully acknowledge the support of the Whitaker Foundation and the John Locke, Jr, Charitable Foundation of Seattle, Wash. We thank Jason Lingel and Marilyn Osterman for assistance in manuscript preparation.
Davies MJ, Thomas AC. Plaque fissuring: the cause of acute myocardial infarction, sudden ischemic death, and crescendo angina. Br Heart J. 1985;53:363-373.
Davies MJ. A macro and micro view of coronary vascular insult in ischemic heart disease. Circulation. 1990;82:1138-1146.
Davies MJ, Richardson PD, Woolf N, Katz DR, Mann J. Risk of thrombosis in human atherosclerotic plaques: role of extracellular lipid, macrophage, and smooth muscle cell content. Br Heart J. 1993;69:377-381.
Falk E. Why do plaques rupture? Circulation. 1992;86(suppl III):III-30-III-42.
Stary HC. The sequence of cell and matrix changes in atherosclerotic lesions of coronary arteries in the first forty years of life. Eur Heart J. 1990;11:3-19.
Lendon CL, Davies MJ, Born GV, Richardson PD. Atherosclerotic plaque caps are locally weakened when macrophage density is increased. Atherosclerosis. 1987;87:87-90.
Nikkari ST, O’Brien KD, Ferguson M, Hatsukami TS, Welgus HG, Alpers CE, Clowes AW. Interstitial collagenase (MMP-1) expression in human carotid atherosclerosis. Circulation. 1995;92:1393-1398.
Brown BG, Zhao XQ, Sacco DE, Albers JJ. Lipid lowering and plaque regression: new insights into prevention of plaque disruption and clinical events in coronary disease. Circulation. 1993;87:1781-1791.
Katz SS, Shipley GG, Small DM. Physical chemistry of the lipids of human atherosclerotic lesions: demonstration of a lesion intermediate between fatty streaks and advanced plaques. J Clin Invest. 1976;58:200-211.
Guyton JR, Klemp KF. Development of the atherosclerotic core region. Arterioscler Thromb. 1994;14:1305-1314.
Small DM, Shipley GG. Physical-chemical basis of lipid deposition in atherosclerosis. Science. 1974;185:222-229.
Small DM. Progression and regression of atherosclerotic lesions. Arteriosclerosis. 1988;8:103-129.
Yuan C, Murakami JW, Hayes CE, Tsuruda JS, Hatsukami TS, Wildy KS, Ferguson MS, Strandness DE. Phased-array magnetic resonance imaging of the carotid artery bifurcation: preliminary results in healthy volunteers and a patient with atherosclerotic disease. J Magn Reson Imaging. 1995;5:561-565.
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.
Pearlman JD, Southern JF, Ackerman JL. Nuclear magnetic resonance microscopy of atheroma in human coronary arteries. Angiology. 1991;42:726-733.
Shahrokh Z, Rapp JH, Shohet SB, Kaufman L. Characterization of atherosclerotic plaque by magnetic resonance imaging. Frontiers in Cardiovascular Imaging. New York, NY: Raven Press; 1993:301-320.
Brown BG, Fry DL. The fate and fibrogenic potential of subintimal implants of crystalline lipid in the canine aorta: quantitative histological and autoradiographic studies. Circ Res. 1978;43:261-273.
Murakami JW, Hayes CE, Weinberger ED. Intensity correction of phased-array coil image. Magn Reson Med. 1996; 35:585-590.
Elster AD. Questions and Answers in Magnetic Resonance Imaging. Mosby-Year Book Inc; 1994.
Small DM. Sterols and sterol esters. In: Small DM, ed. Handbook of Lipid Research, 4: The Physical Chemistry of Lipids: From Alkanes to Phospholipids. New York, NY: Plenum Press; 1986:395-425.