Intravascular Magnetic Resonance Imaging of Aortic Atherosclerotic Plaque Composition
Abstract Magnetic resonance imaging (MRI) may be an excellent tool to define atherosclerotic plaque composition, but surface MRI (SMRI) suffers from a low signal-to-noise ratio and low resolution of arterial images. Intravascular MRI (IVMRI) represents a potential solution for acquiring high-quality in vivo images of atherosclerotic plaques. Isolated segments of 11 thoracic human aortas obtained at autopsy were imaged by IVMRI using an intravascular receiver catheter coil designed and built at our institution. Images obtained by IVMRI were compared with corresponding images obtained by SMRI and with histopathological aortic cross sections. The intensity of intimal thickness and plaque components was graded by IVMRI and histopathology using a score of 1 for mild, 2 for moderate, and 3 for severe intensity. IVMRI had an agreement of 75% with histopathology in fibrous cap grading (37.5% expected, κ=0.60, P<0.001) and of 74% in necrotic core grading (39% expected, κ=0.57, P<0.001). Intraplaque calcification was correctly graded by IVMRI in six of the eight plaques in which histopathology recognized calcium. The analysis of intimal thickness showed 80% agreement between IVMRI and histopathology (52% expected, κ=0.59, P<0.001). IVMRI image features were similar to those of SMRI. In addition, IVMRI accurately determined atherosclerotic plaque size in comparison with histopathology and SMRI (slope=1.25 cm2, r=0.99, P<0.001 for luminal area by IVMRI vs histopathology; slope=0.97 cm2, r=0.996, P<0.001 for luminal area by IVMRI vs SMRI). IVMRI has the potential to provide important prognostic information in patients with atherosclerosis because of its ability to accurately assess both plaque composition and size.
- Received February 14, 1997.
- Accepted September 18, 1997.
The assessment of atherosclerotic plaque composition has significant prognostic implications for patients with atherosclerosis, because it is one determinant of the susceptibility to plaque disruption and resultant clinical outcomes in patients with atherosclerotic disease. The amount of intraplaque lipid content and the thickness of the fibrous cap are closely related to the vulnerability of a given plaque to rupture,1 2 with resultant untoward clinical outcomes including unstable angina, myocardial infarction, and stroke. Other intraplaque elements may also contribute to plaque disruption. The extent and location of intraplaque calcification are also related to atherosclerotic plaque vulnerability.3 4 5 Hemorrhage resulting from rupture of vasa vasorum within the plaque, causing increased intraplaque pressure, may also be a responsible trigger.6
Current clinical magnetic resonance techniques do not provide a sufficient signal-to-noise ratio to achieve high-resolution arterial images in vivo because of the distance between the imaging coil and the vessel imaged. MRI has been shown to demonstrate plaque structure in detail using image contrast from different intraplaque components.7 8 9 10 However, these studies applied a surface coil directly to the vessel, which is not feasible in vivo. A practical solution for this problem is to acquire the image from inside the vessel using an intravascular catheter. The first successful attempt at using intravascular magnetic resonance was performed by Kantor et al11 to obtain nuclear magnetic resonance spectroscopy from the right and left ventricles of canine hearts. The same principle was later used as an imaging technique in the study of atherosclerotic plaques from inside large vessels, both ex vivo12 13 14 and in vivo15 16 with animal models, demonstrating that the signal-to-noise ratio can be improved and that high-resolution images of arteries and veins can be obtained by using an intravascular catheter coil. It should also be noted that IVMRI of small arteries in vivo is feasible, as illustrated by our experimental studies in the rabbit model.17
Although the ability of IVMRI to provide good-quality images has been demonstrated, detailed determination of plaque composition and size by IVMRI has not been attempted. In the present study, an intravascular coil was used to determine plaque composition and size in isolated human aortae.
Aortic arterial segments were removed from 11 individuals aged 44 to 85 years (mean±SD, 65±15 years) at the time of autopsy. Approximately 10 cm of thoracic aorta was excised, and both ends of the vessel were occluded with a large cork. The caudal cork had an orifice to allow introduction of the IVMRI catheter. The aorta was taped inside a plastic container filled with water at room temperature.
The experiments were performed on a GE Signa 1.5 T scanner. The receiver catheter coil, designed and built at our institution, was long, narrow, and flexible. The coil was a segment of standard wire with conductors (diameter=3 mm or 9F) shorted at one end and with the other end used as the coil terminal. It was tuned with small, fixed ceramic chip capacitors (1.5×1.5×1.4 mm) sealed with plastic dip. A 50-Ω coaxial cable was used to transmit the magnetic resonance signal to the preamplifier. For decoupling the transmitter (body) coil and the receiver (catheter) coil, a PIN diode at the end of the coaxial cable turned the coil off during radiofrequency transmission via a DC pulse supplied by the scanner. The catheter coil was described in detail in a previous article.14
The MRI standard protocol lasted 30 minutes and consisted of IVMRI and SMRI axial images of the artery. The plastic container with aorta and the coil inside was placed in the scanner and aligned with the main magnetic field. After coronal scout images, axial images were obtained using the following spin-echo imaging protocol: image matrix, 256×256; pixel dimensions, 0.27×0.27 mm; slice thickness, 3 mm; TR, 1500milliseconds; two echoes; TE, 17 and 80 milliseconds; acquisition time, 12 minutes, 51 seconds; 2 NEX; and 14 slices. Proton density (TR=1500 milliseconds/TE=17 milliseconds) and T2-weighted (TR=1500 milliseconds/TE=80 milliseconds) were the two types of images used. Intravascular images were acquired from the catheter coil placed inside the vessel along its length and without moving it. After IVMRI, the catheter coil was then removed and the surface coil placed under the plastic container for SMRI. SMRI images were obtained by placing a 75-mm-diameter surface coil directly under the plastic container, using exactly the same parameters as used to acquire the intravascular images.
After MRI acquisition, the aorta was placed in 4% formaldehyde solution for at least 48 hours. The fixed aorta was cut into cross-sectional slices that matched the location (every 5 mm) of the images acquired by IVMRI. The aortic cross-sectional slices were stained with Oil Red O, embedded in paraffin, and cut into sections 5 μm thick, which were processed for hematoxylin-eosin, Verhoeff-van Gieson, and Masson staining.
Histopathology was used for two specific purposes: plaque composition analysis and plaque size measurements. Photomicrographs of the hematoxylin-eosin—stained sections of were digitized onto compact discs and analyzed using National Institutes of Health Image software to determine the degree of aortic atherosclerotic stenosis. Verhoeff-van Gieson and Masson stains were performed using standard methods to enable recognition of elastic components and to better differentiate the aortic wall layers in the plaque composition analysis.
Image Data Analysis
Arterial images were acquired by IVMRI and SMRI every 5 mm from the caudal to the cranial end of the aortic segment. Histopathological sections were also obtained every 5 mm of the aorta. In this manner, the sites imaged by the MRI methods corresponded to the location of histopathological cross sections. Anatomic landmarks were also used to carefully fine-tune the match between IVMRI and histopathology cross sections in the atherosclerotic plaque composition analysis.
The plaque composition analysis consisted of two parts. In the first, IVMRI images were compared with histopathology slides, and image features of individual plaque components and arterial wall layers by IVMRI were determined. Image characteristics were also expressed in terms of T2 relaxation times for each component. In the second part, the ability of IVMRI to estimate intimal thickening, fibrous cap thickness, necrotic core size, and the extent of intraplaque calcification was evaluated by grading these structures (1=mild, 2=moderate, 3=severe) and comparing these scores with those obtained from histopathological slides carefully matched to the MRI images. One investigator graded the severity of the various components of the atherosclerotic plaque by histopathology and another investigator by IVMRI. Each was blinded to the results of the other modality.
In the plaque size analysis, MRI and histopathology measurements were compared by careful registration of corresponding images as described above. CSA was defined as the total area circumscribed by the intima/media or plaque/media interface, and ILA was defined as the total area circumscribed by the intima/lumen interface. The degree of atherosclerotic stenosis was calculated as (CSA−ILA)/CSA. The measurements were performed by two different investigators. The arithmetic mean of the two measurements of ILA or CSA performed by different investigators at each aortic site was taken as the final value. Interobserver variation was also calculated.
The κ test was used to assess agreement between the IVMRI and histopathology scores utilized to evaluate plaque composition and intimal thickening. One-way analysis of variance with Bonferroni correction was used to compare T2 measurements from different plaque and arterial wall components, which were expressed as the mean±SE. In the atherosclerotic plaque size analysis, agreement among the methods was assessed by limits of agreement as illustrated on Bland-Altman plots, and simple linear regression was used to correlate quantitative measurements obtained by IVMRI, SMRI, and histopathology. The linear regression equation was forced through the origin (0,0) to identify differences in measurements based on the slope value. Limits of agreement were also used to assess interobserver variation of IVMRI and SMRI. For all analyses a value of P<0.05 was required for statistical significance.
Aortic Arterial Wall
The best contrast between arterial layers was achieved by T2-weighted images. IVMRI identified the aortic intima as a very thin, dark membrane, better recognized when thickened and identified as a low signal inner layer on the T2-weighted images (Fig 1⇓). The arterial media could be systematically recognized by IVMRI and appeared on T2-weighted images as a high-signal (bright) layer. The arterial adventitia corresponded to a low-signal (dark) outer layer and was also easily recognized because of its contrast with the medium.
The intima and adventitia had low T2 values (mean±SE) (intima=46.7±2 milliseconds, n=33; adventitia=52.2±9 milliseconds, n=20), while the media had high T2 values (mean=91.2±10 milliseconds, n=60) (Fig 2⇓). T2 values from the arterial media were statistically different (P<0.05) from the intima and adventitia values.
Atherosclerotic Plaque Components
The fibrous cap was recognized as a dark structure covering the atherosclerotic plaque, while the necrotic core appeared as a bright nucleus under the fibrous cap (Fig 3⇓). The fibrous cap had a low T2 value (mean±SE=49.2±4 milliseconds, n=16), while the highest T2 values were obtained from the necrotic core and the medial elastic layer (75.6±9 milliseconds, n=29, and 91.2±10 milliseconds, n=60, respectively). Both T2 values from necrotic core and media were significantly greater (P<0.05) than the ones from the adventitia, fibrous cap, and intima reported above (Fig 2⇑). There was no statistical difference between media and necrotic core T2 values.
Intraplaque calcification appeared as a signal void on IVMRI, because calcified regions have low proton densities (Fig 4⇓, A). All intraplaque features obtained by MRI performed with a surface coil applied to the vessel (Fig 4⇓, B) were similar to the IVMRI features described above.
The ability of IVMRI to estimate the magnitude of individual atherosclerotic plaque components and arterial intimal thickness was also compared with that of histopathology using a score of 1 to 3 (1=mild, 2=moderate, 3=severe). Sixty IVMRI-histopathology–matched cross-sectional images were analyzed. The analysis of these images revealed 24 atherosclerotic plaques whose composition was compared by IVMRI and histopathology.
IVMRI recognized a fibrous cap in 20 of the 24 plaques in which this structure was identified by histopathology. The 20 fibrous caps were graded by both methods, and the agreement between the two was 75% (37.5% expected, κ=0.60, P<0.001).
A necrotic core was recognized by histopathology in 23 of the 24 plaques evaluated. IVMRI detected this finding in all 23 plaques but incorrectly recognized a necrotic core in one in which this feature was not recognized by histopathology. In grading the extent of intraplaque lipid accumulation, histopathology and IVMRI had an agreement of 74% (39% expected, κ=0.57, P<0.001).
Intraplaque calcification was recognized by histopathology in 8 of the 24 plaques analyzed. IVMRI correctly identified such calcifications in 7 of them. In addition, IVMRI correctly graded calcification in all cases compared with histopathology: 6 as severe and 1 as moderate calcification.
Histopathology showed intimal thickening in 46 of the 48 sections analyzed for this purpose. IVMRI correctly recognized thickening in all 46 and incorrectly recognized it as present in one of the two cases where it was not recognized by histopathology. κ analysis showed 80% agreement between IVMRI and histopathology grades of thickness (52% expected, κ=0.59, P<0.001).
Atherosclerotic Plaque Size Determination
Interobserver variations in SMRI and IVMRI measurements were assessed by limits of agreement. The means of the differences between the two observers’ measurements of CSA and ILA by IVMRI were −0.14 cm2 and −0.03 cm2, while the SDs of the differences were 0.15 cm2 and 0.08 cm2, respectively. The means of the differences between the two observers’ measurements of CSA and ILA by SMRI were −0.12 cm2 and −0.02 cm2, while the SDs of the differences were 0.13 cm2 and 0.09 cm2, respectively.
Aortic atherosclerotic plaque size measured by IVMRI was compared with measurements derived from SMRI images and histopathology. Table 1⇓ summarizes limits of agreement analysis of the comparison between IVMRI and SMRI, showing a high level of agreement. The SDs of the differences between IVMRI and SMRI measurements of CSA and ILA were 0.22 cm2 and 0.25 cm2, respectively. Linear regression analysis was performed between SMRI and IVMRI for CSA (slope=0.97, r=0.998, P<0.001), ILA (slope=0.97, r=0.996, P<0.001, Fig 5⇓, A), and percent stenosis (slope=0.98, r=0.94, P<0.001).
Table 2⇓ summarizes limits of agreement analysis of the comparison between IVMRI and histopathology, showing that absolute values of CSA and ILA were systematically lower by histopathology than by IVMRI. The SDs of the differences between IVMRI and histopathology measurements of CSA and ILA were 0.40 cm2 and 0.36 cm2, respectively, higher than in the comparison with SMRI. This discrepancy was caused by the staining process required for histopathological examination, which causes systematic shape changes and shrinkage of aortic tissue.10 Slope values obtained by linear regression forcing the analysis through zero indicates that IVMRI measures of CSA (slope=1.26, r=0.99, P<0.001) and ILA (slope=1.25, r=0.99, P<0.001, Fig 5⇑, B) were 26% and 25% greater than the histopathology measures, respectively. Percent stenosis analysis showed that the slope=0.94, r=0.90, and P<0.001.
Our study demonstrates the ability of IVMRI to characterize atherosclerotic plaque composition, revealing a significant advantage of this method over currently used arterial imaging techniques. We also found that atherosclerotic burden in terms of plaque size can be accurately quantified by IVMRI.
Although prior studies7 8 9 10 have indicated that MRI obtained with a surface coil directly applied to the vessel provides detailed characterization of atherosclerotic plaques, the distance between coil and vessel with this technique is too large, except for superficial arteries,18 to achieve high-resolution images in vivo. In the present study, we used an intravascular catheter receiver coil,12 13 14 15 16 17 which represents a practical solution for achieving detailed arterial imaging in vivo, to differentiate fibrous cap from lipid core within an atherosclerotic plaque. By using a scoring system to estimate the extent of abnormality, we graded intimal thickness and intraplaque components by IVMRI and demonstrated a high level of agreement with histopathology.
We observed that the fibrous cap appears as a short T2 plaque component and the necrotic core as a long T2 plaque component, in agreement with the findings of Yuan et al.7 Conversely, Martin et al9 reported the fibrous cap to have a long T2 and the necrotic core a short T2 plaque component. As discussed by Berr et al,19 the appearance of intraplaque lipids in MRI images depends on their physical state, which can vary with factors such as temperature, age of the plaque, and time of storage. Also, the composition of the fibrous cap, particularly the amount of collagen, influences the characteristics of the images. In addition, differences in pulse sequences, field strength, and other image parameters may explain differences in image contrast.
The potential limitations of this data also merit consideration. The fixation process should ideally have been performed under systemic pressure applied to the vessel to avoid changes in dimensions. To assess the effect of our fixation process on vessel dimensions, we imaged, after fixation, 7 of the 11 aortae (a total of 56 segments) by surface MRI and compared these dimensions with the ones from surface MRI before fixation and histopathology. We concluded that, in our experiments, the changes in vessel dimensions by the histological process occurred mainly after the fixation phase, because measurements obtained by SRMI before fixation were not different than the ones obtained after fixation. However, there was a clear difference between measurements obtained by SMRI after fixation and histopathology. Therefore, the fact that fixation was performed without pressurization does not significantly decrease the precision of our measurements.
The matching between MRI images and histopathology sections based on metric criteria is limited, because final histopathology paraffin sections were 5 μm thick, while MRI slices were 3 mm thick. To obtain more precise matching for the qualitative analysis, the images were visually compared, and anatomic landmarks were used to check the matches.
The 15-minute acquisition time reported in our methods is relatively long, because we used spin-echo as a pulse sequence. Fast pulse sequences, such as fast spin-echo, GRASS, and echo-planar could have beenused, which would have optimized acquisition time without decreasing image quality. For example, our group used spoiled GRASS to acquire IVMRI of rabbit aortae in vivo.17
Because fibrous cap thickness and lipid core size represent the main factors that determine plaque vulnerability to rupture,1 2 it may be possible to use the IVMRI technique to image coronary arteries in patients with ischemic heart disease, thus providing prognostic information that may be of more value than merely the number of lesions and degree of luminal obstruction, as currently detected by angiography. In addition, the extent of intraplaque calcification3 4 5 and hemorrhage6 may also determine the evolution of atherosclerotic plaques, and IVMRI may also be able to recognize and characterize both of these plaque alterations.
The current most commonly used method to evaluate atherosclerotic plaques, x-ray angiography, has the ability to demonstrate plaque site and degree of stenosis but is not able to accurately assess the predisposition of a given plaque to rupture. The reason is the limited capability to determine plaque composition, because this method does not provide direct visualization of the arterial wall. Angiography also has limitations in quantifying plaque size, because it provides images only of the arterial lumen. Thus, plaque stenosis is commonly underestimated if adjacent arterial segments are also involved with disease or when compensatory dilatation of an involved arterial segment is present.20 21 Percutaneous transluminal angioscopy, which provides direct visual access to the arterial endothelial surface, appears to be more sensitive than x-ray angiography in detecting the degree of luminal stenosis but can examine only the surface of the lesion, therefore providing limited information on plaque composition.22 Intravascular ultrasound has been extensively used in clinical practice and clinical investigation to characterize atherosclerosis. It is more precise than x-ray angiography in quantifying atherosclerotic plaque size but is limited in the evaluation of plaque composition because of poor contrast resolution between different intraplaque components.23 24
MRI was originally developed to study the brain and static body organs because it relied upon averaging signals from a large number of radiofrequency excitations over time. Recently, however, the development of fast imaging technology25 26 to study the cardiovascular system has made MRI a powerful tool for clinical investigation.27 28 29 30 31 The potential of integrating recent MRI developments with open magnets, which are designed to allow invasive procedures guided by MRI, creates the possibility of using IVMRI in humans to study atherosclerotic disease. In this regard, a loopless catheter antenna less than 2.5F in diameter has been developed17 that might be used to image coronary arteries. This catheter antenna has been tested in vivo to acquire images of rabbit aortae, which are slightly larger in diameter than human coronary arteries (Fig 6⇓).
In conclusion, IVMRI provides detailed characterization of intraplaque components, including the extent of intraplaque lipid accumulation and fibrous cap thickness. We believe this new IVMRI technique has the potential to provide prognostic information as well as to assess the response to different therapeutic interventions in patients with atherosclerotic disease.
Selected Abbreviations and Acronyms
|IVMRI||=||intravascular magnetic resonance imaging|
|MRI||=||magnetic resonance imaging|
|NEX||=||number of excitations|
|SMRI||=||surface magnetic resonance imaging|
This work was supported by a Whitaker Foundation biomedical research grant, Grant-in-Aid 92–10-26–01 from the American Heart Association, and Grant RO1-HL-43722 from the National Heart, Lung, and Blood Institute, National Institutes of Health. Special thanks to Mary McAllister for her help in manuscript preparation, to Bradley D. Bolster, Jr, for providing the animal in vivo image, and to Dr James Tonascia for his crucial help with the statistical analysis.
Falk E, Shah PK, Fuster V. Coronary plaque disruption. Circulation.. 1995;92:657–671.
Demer L. Effect of calcification on in vivo mechanical response of rabbit arteries to balloon dilatation. Circulation.. 1991;83:2083–2093.
Merickel MB, Berr S, Spetz K, et al. Non-invasive evaluation of atherosclerosis utilizing 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.
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.
Thomas AC, Davies MJ, Dilly S, Dilly N, Franc F. Potential errors in the estimation of coronary arterial stenosis from clinical arteriography with reference to the shape of the coronary arterial lumen. Br Heart J.. 1986;55:129–139.
Nissen SE, Gurley JC, Grines CL, et al. Intravascular ultrasound assessment of lumen size and wall morphology in normal subjects and patients with coronary artery disease. Circulation.. 1991;84:1087–1099.
Wendland MF, Saeed M, Higgins CB. Strategies for differential enhancement of myocardial ischemia using echoplanar imaging. Invest Radiol. 1991; 26:S236–S238.
Lima JAC, Judd RM, Bazille A, Schulman SP, Atalar E, Zerhouni EA. Regional heterogeneity of human myocardial infarcts demonstrated by contrast-enhanced MRI: potential mechanisms. Circulation.. 1995;92:1117–1125.
Clarke GD, Eckels R, Chaney C. Measurement of absolute epicardial coronary artery flow and flow reserve with breath-hold cine phase-contrast magnetic resonance imaging. Circulation.. 1995;91:2627–2634.