Classification of Human Coronary Atherosclerotic Plaques Ex-Vivo with T1, T2 and Ultra-short TE MRI

Abstract

Objectives

To determine whether the classification of human coronary atherosclerotic plaques with T1, T2 and ultrashort TE (UTE) MRI would correlate well with atherosclerotic plaque classification by histology.

Background

MRI has been extensively used to classify carotid plaque but its ability to characterize coronary plaque remains unknown. In addition, the detection of plaque calcification by MRI remains challenging. Here we used T1, T2 and UTE MRI to evaluate atherosclerotic plaques in fixed post-mortem human coronary arteries. We hypothesized that the combination of T1, T2 and UTE MRI would allow both calcified and lipid-rich coronary plaques to be accurately detected.

Methods

28 plaques from human donor hearts with proven coronary artery disease were imaged at 9.4 Tesla with a T1 weighted 3D FLASH sequence (250 um resolution), a T2 weighted Rare sequence (in plane resolution 0.156mm), and an UTE sequence (300um resolution). Plaques showing selective hypointensity on T2 weighted MRI were classified as lipid-rich. Areas of hypointensity on the T1 weighted images but not the UTE images were classified as calcified. Hyperintensity on the T1 weighted and UTE images was classified as hemorrhage. Following MRI, histological characterization of the plaques was performed with a pentachrome stain and established AHA criteria.

Results

MRI showed high sensitivity and specificity for the detection of calcification (100% and 90%) and lipid-rich necrotic cores (90% and 75%). Only two lipid-rich foci were missed by MRI, both of which were extremely small. Overall, MRI based classification of plaque was in complete agreement with the histological classification in 22/28 cases (weighted κ =0.6945, p<0.0001).

Conclusions

The utilization of UTE MRI allows plaque calcification in the coronary arteries to be robustly detected. High-resolution MRI with T1, T2 and UTE contrast enables accurate classification of human coronary atherosclerotic plaque.

Introduction

Coronary atherosclerosis remains the leading cause of morbidity and mortality in industrialized countries (1). Most acute coronary events can be linked to the rupture of a coronary atherosclerotic plaque, resulting in the formation of a luminal thrombus and possibly occlusion of the vessel (2). Post mortem studies have identified features specific to plaques vulnerable to rupture such as a large lipid core and a thin fibrotic cap separating the core from the vessel lumen (3–5). Invasive techniques such as intravascular ultrasound and optical coherence tomography have been used to study plaque morphology in vivo (6,7), but are not appropriate for large-scale screening or serial follow-up studies. Significant interest thus exists in the development of noninvasive imaging techniques to characterize atherosclerotic coronary plaques and assess their risk for rupture and complication.

Computed tomography (CT) of the coronary arteries allows the vessel wall to be imaged with high spatial resolution, but its soft tissue characterization is limited (8,9). The experience with MRI characterization of carotid plaque is now extremely extensive (10). Multicontrast MRI is able to detect carotid plaque lipid content, hemorrhage, neovascularization and fibrous cap thickness (11–15). Recently, ultra-short echo time (UTE) MRI has been used to detect and quantify plaque calcification in the carotids (16,17). MRI characterization of coronary plaque, however, is in its nascency. Unlike MRI of the carotids, MRI of the coronary wall currently lacks the spatial resolution needed for the detailed characterization of plaque morphology.

Despite recent developments in high-field cardiovascular MRI (18–20), the ability of high-resolution MRI to accurately characterize human coronary atherosclerotic plaque remains poorly defined. Only one study to date has examined the correlation between human coronary artery plaque by MRI and histology (21), and this study was hindered by the absence of a sequence to specifically detect plaque calcification. The aim of the current study was to use high-resolution T1, T2 and ultrashort TE (UTE) MRI to classify human coronary atherosclerotic plaques. We hypothesized that the combination of T1, T2 and UTE MRI would allow potentially vulnerable lipid-rich plaques to be distinguished from more stable fibrocalcific plaques with a high degree of accuracy. We further hypothesized that the classification of human coronary plaque morphology with T1, T2 and UTE MRI would correlate strongly with the histological plaque classification, and further validate the use of MRI to assess plaque morphology in human coronary arteries.

Methods

Preparation and imaging of the donor hearts

Human hearts (n=3), rejected for transplantation because of significant coronary artery disease (CAD), were obtained from the International Institute for the Advancement of Medicine (IIAM, Jessup, PA). The LAD ostium was selectively cannulated with a plastic luer, which was connected to a pressure-perfusion system. 10% buffered neutral formalin solution was infused at 130 mmHg for 30 minutes for tissue fixation. A block of myocardium, including the LAD and its proximal side branches was then excised from the heart for MRI. All studies were performed in accordance with a protocol approved by local Institutional Review Board (IRB).

MRI images were acquired on a 9.4T horizontal bore magnetic resonance scanner (Biospec, Bruker, Billerica, MA, USA) using a transmit-receive birdcage coil. The specimens were immersed in a fluorcarbon matching medium with the same magnetic susceptibility as tissue but no proton MR signal, thus providing a signal and artifact free background (Fomblin, Ausimont, NJ). Specimens were imaged with T1, T2 and UTE sequences. T1-weighted imaging was achieved with a 3D FLASH sequence using the following parameters: FOV 48×36×64mm³, image matrix 192×144×256 (0.250mm isotropic resolution), TR=30ms, TE=2.5ms, flip angle 45 degrees, fat suppression, 1 average. T2-weighted imaging was performed using a 2D RARE sequence with the following parameters: FOV 40×30mm², image matrix 256×192 (in plane resolution 0.156mm), slice thickness 0.4mm, TR=3000ms, TE=40ms, echo train 16, 16 averages. UTE imaging was performed using a 3D radial sequence with the following parameters: FOV 51×26×60mm³, image matrix 192×192×192 (resolution 0.267×0.133×0.312mm³) projections 75546, TR=8ms TE=20μs, fat suppression, 1 average.

Histopathology and Image Coregistration

The histological processing and analysis was performed at a pathology institute specialized in cardiovascular histopathology (CV Path Laboratory, Maryland, MD). The LAD specimens were embedded en bloc in paraffin. Cross-sections were acquired in 1-mm increments (6μm slice thickness) and Movat’s pentachrome staining was used according to the standard techniques (22). Analysis of the sections was performed by an experienced pathologist blinded to the MRI data. Main plaque components such as lipid-rich necrotic core and calcification were reported. Further classification of the coronary plaques was performed according to the histological criteria established by the AHA (3).

Manual co-registration of the MRI data and the digitized histological sections was performed by an experienced reader, who was not participating in the further analysis. Multi-planar reformats of the MRI images perpendicular to the vessel centerline were prepared in 1-mm increments to match to the histological slides. Anatomical features such as distance from the coronary ostia, presence of side-branches or vessel bifurcation, vessel size and shape, and plaque morphology were used to perform the rigid co-registration. The T1, T2 and UTE MRI images were coregistered using a freeware Dicom reader (OsiriX, Geneva, Switzerland).

Image Analysis

Plaque classification was performed by three independent observers blinded to the histological plaque classification. In cases of disagreement a consensus was reached in a subsequent reading. The MRI signal and contrast characteristics used to identify the relevant plaque components are shown in Table 1. Areas of plaque showing signal hypointensity on the T1 but not the UTE images were classified as calcified. Areas showing signal hypointensity on the T2 images were classified as lipid-rich necrotic core. Hemorrhage was identified by hyperintensity on the T1 and UTE images and regions that were isointense on all sequences were classified as fibrotic (Table 1). Signal intensities were interpreted with the adjacent myocardium as a reference. The MRI classification system used (Table 2) was based on that proposed by Cai et al. for carotids (23), but was adopted to include the assessment of plaque calcification using the UTE technique (16,17). The details of the classification scheme used are provided in Table 2 and in the supplemental methods.

Table 1.

Predicted signal intensities of plaque components by MRI

	Ca	LRNC	Fibrotic	Hemorrhage	* Tissue Tear/Disruption
T1W	Hypo	Iso	Iso	Hyper	Hypo
T2W	Hypo	Hypo	Iso	Hyper/Hypo	Hypo
UTE	Iso	Iso	Iso	Hyper	Hypo

Ca = calcification, LRNC = lipid-rich necrotic core, Hypo = hypointense, Iso = isointense, Hyper = hyperintense.

^*Results in the local accumulation of the fluorocarbon matching medium.

Table 2.

Scheme for classification of atherosclerotic plaques

Conventional AHA Classification 4	Modified AHA classification for MRI 14	Modified AHA classification for MRI adapted for UTE imaging
Type I: initial lesion with foam cells	Type I–II: near-normal wall thickness, no calcification	Type I–II: near normal wall thickness, no calcification
Type II: fatty streak with multiple foam cell layers
Type III: preatheroma with extracellular lipid pools	Type III: diffuse intimal thickening or small eccentric plaque with no calcification	Type III: diffuse intimal thickening of small eccentric plaque with no calcification
Type IV: atheroma with a confluent extracellular lipid core	Type IV–V: plaque with a lipid or necrotic core surrounded by fibrous tissue with possible calcification	Type IV–VA: plaque with lipid or necrotic core surrounded by fibrous tissue without calcium
Type V: fibroatheroma		Type IV–VB: plaque with lipid or necrotic core surrounded by fibrous tissue with calcium
Type VI: complex plaque with possible surface defect, hemorrhage, or thrombus	Type VI: complex plaque with possible surface defect, hemorrhage, or thrombus	Type VI: complex plaque with possible surface defect, hemorrhage, or thrombus
Type VII: calcified plaque	Type VII: calcified plaque	Type VII: calcified plaque
Type VIII: fibrotic plaque without lipid core	Type VIII: fibrotic plaque without lipid core and with possible small calcifications	Type VIIIA: fibrotic plaque without lipid core and without calcification
		Type VIIIB: fibrotic plaque without lipid core and with calcium

Statistical analysis

Continuous variables with normal distributions are expressed as mean ± standard deviation, while categorical variables are given as counts and percentages. Inter-observer agreement for plaque classification was determined by using the kappa statistic as reported previously (24). Using histology as the gold standard, the sensitivity, specificity, positive and negative predictive values (plus binominal 95% confidence intervals) of MRI for the detection of calcium and lipid-rich necrotic core were calculated. The agreement between MRI classification and the gold standard of histology was calculated using weighted kappa and interpreted as follows: excellent agreement: κ > 0.81; good agreement: κ= 0.61–0.80; and moderate agreement: κ=0.41–5.60. Statistical analysis was performed using commercially available software (SAS, version 9.2 SAS Institute Inc., Cary, NC, USA). A p-value of <0.05 was considered significant.

Results

28 plaque cross sections could be co-registered with corresponding MRI datasets (T1, T2, UTE) and were used for the analysis. Multicontrast MRI had an excellent sensitivity (100%), specificity (90%) and both positive (80%) and negative (100%) predictive values for the identification of calcium (Table 3). The sensitivity, specificity, positive and negative predictive values for lipid-rich necrotic core recognition with MRI were 90%, 75%, 90% and 75 %, respectively (Table 3). MRI failed to identify lipid-rich necrotic areas in two plaques, however, both these lesions were classified histologically as early lesions with only a very small lipid-rich necrotic cores.

Table 3.

Detection of Calcification (Ca) and Lipid-Rich Necrotic Core (LRNC) by MRI

Calcification (Ca)
	Ca on Histology	No Ca on Histology	n
Ca on MRI	8	2	10
No Ca on MRI	0	18	18
n	8	20	28
Sensitivity: 8/8 = 100% (95%CI, 63–100%)		PPV: 8/10 = 80% (95%CI, 44–98%)
Specificity: 18/20 = 90% (95%CI, 68–99%)		NPV: 18/18 = 100% (95%CI, 82–100%)
Lipid-Rich Necrotic Core (LRNC)
	LRNC on Histology	No LRNC on Histology	n
LRNC on MRI	18	2	20
No LRNC on MRI	2	6	8
n	20	8	28
Sensitivity: 18/20 = 90% (95%CI, 68–99%)		PPV: 18/20 = 90% (95%CI, 68–99%)
Specificity: 6/8 = 75% (95%CI, 35–97%)		NPV: 6/8 = 75% (95%CI, 35–97%)

PPV = positive predictive value; NPV = negative predictive value, Ca = calcification

Table 4 shows the comparison of plaque classification by MRI and histology. Complete agreement was seen in 22/28 cases. Type I–II lesions were identified correctly by MRI in all cases. Two sections with Type III plaques by histology were misclassified by MRI as having small lipid-rich necrotic cores. It should be noted that prior to the consensus discussion, 2/3 readers had classified these plaques as Type III by MRI as well. Among Type IV–V lesions, 2 plaques were falsely classified as calcified by MRI. Overall only 6/28 lesions were misclassified by MRI, resulting in a good correlation between MRI and histology (κ value of 0.69). Excellent inter-observer agreement was found for the detection of plaque components (calcification and lipid-rich necrotic core) by MRI with a κ value of 0.80 and 0.82, respectively (p<0.0001). The inter-observer correlation for overall MRI plaque classification was also high (κ=0.78, p<0.0001).

Table 4.

Classification of coronary plaques by MRI vs. histology

	Histology
MRI		Type I–II	Type III	Type IV–V (A)	Type IV–V (B)	Total
	Type I–II	2	…	…	…	2
	Type III		4	…	…	4
	Type IV–V (A)	…	2	9	…	11
	Type IV–V (B)	…	…	2	7	9
	Type VIII (A)	…	…	1	…	1
	Type VIII (B)	…	…	…	1	1
	Total	2	6	12	8	28

Discussion

The utility of cardiac MRI in the classification of human coronary atherosclerotic plaque remains poorly defined. Moreover, coronary atherosclerotic plaques frequently calcify and any MRI classification scheme must therefore be able to reliably detect plaque calcification. Here we show that a combination of T1, T2 and UTE MRI can robustly classify human coronary atherosclerotic plaques, including lipid-rich and calcified lesions. We show that plaque classification with this multicontrast approach correlates very strongly with plaque classification by histology. To the best of our knowledge, this is the first use of a UTE-based triple-contrast approach (T1, T2, UTE) in the evaluation of human coronary atherosclerosis.

Histological and invasive studies have shown that the majority of acute coronary syndromes result from rupture of advanced atherosclerotic plaques (3–5). These plaques typically have a large lipid-rich necrotic core covered by a thin fibrous cap (4). The histological threshold used to define a thin cap is 65 um (4), and is beyond the resolution of all non-invasive imaging techniques. Nevertheless, we show that MRI can accurately differentiate fibrous coronary plaques with no necrotic cores from those containing large lipid-rich necrotic cores. The role of calcification in plaque vulnerability remains unclear. While some have suggested that heavily calcified plaques are more stable (25,26), this is disputed by others (27,28). Regardless, accurate detection of calcification is vital for reliable plaque classification and is made possible by the addition UTE MRI.

The use of multicontrast MRI is well established in the carotid arteries, and several studies have shown its strong correlation with histological analysis of endarterectomy specimens (11,12). The discrimination of lipid-rich necrotic cores in carotid arteries was originally described using T1, T2 and Proton Density (PD) weighted spin echo sequences with fat suppression. These studies report the delineation of lipid-rich necrotic core with a sensitivity of 90% and specificity of 65–74% (11,12). The use of UTE imaging in the carotid arteries ex-vivo has also recently been described (16,17). While the experience in the carotid vascular bed is extremely valuable, a direct correlation with plaque morphology in the coronary arteries cannot be assumed. Direct imaging of plaque morphology with MRI in the coronary arteries is thus required to characterize the disease. The results of our study show that the classification of atherosclerotic plaque using multicontrast MRI produces similar accuracy in the coronary and carotid arteries. Moreover, the incorporation of UTE MRI in our study resulted in a higher specificity for necrotic core detection compared to some of the prior carotid imaging studies (11,12).

Several sequences, routinely performed during in vivo MRI of the carotids, could not be performed in this ex vivo study. The intravenous injection of gadolinium-based contrast agents in the carotid artery can delineate plaque neovascularization and fibrous cap rupture, but could not be performed in our setting (14,29). Although some of the plaques in our study showed features suggesting surface disruption, the accuracy of this finding in fixed ex vivo coronary arteries remains unknown. We chose therefore to limit our classification to the plaque interior, where fixation has been shown not to have an effect on plaque characterization by MRI (30). Intra-plaque hemorrhage produces signal hyperintensity on T1 and off-resonance MRI images (31,32), and signal hypointensity in T2* weighted images (33,34). Plaque hemorrhage can thus be expected to produce marked hyperintensity on UTE images, in which the R2* effects are completely eliminated. No plaques with hemorrhage, however, were encountered in this study and this will need to be further tested in future studies.

Calcified tissues are rigid and hence have extremely short transverse relaxation times (significantly less than 1 ms). Conventional imaging sequences have echo times (TE) greater than 1 ms and are thus unable to detect any signal from the tissues. UTE MRI, however, has a TE in the low microsecond range, and has been used to image bones and carotid plaque calcification (16,17,35). The utilization of UTE imaging in the carotids yielded a sensitivity of 71% and specificity of 96% for the detection of calcium (16). UTE MRI, however, has not been used prior to this in the coronary arteries.

In vivo MRI of the coronary wall has been limited to date to the detection of wall thickening, delayed enhancement and intraplaque hemorrhage (31,36,37). High-resolution ex-vivo imaging of the coronary artery wall has been performed but on a very limited scale (21,30). Itskovich et al imaged ex vivo coronary plaques at 9.4T and used cluster analysis to successfully classify atherosclerotic plaques (21). UTE MRI, however, could not be performed in this study and limited the discrimination of plaque calcification (21). In contrast, the use of UTE MRI in our study allowed plaques with calcified components to be accurately classified. Moreover, we were able to reach higher inter-observer agreement for coronary plaque classification with our technique, than others have reported using T1, T2, PD and 3D imaging in the carotids (38). The field strength used in this study allowed high spatial resolution to be achieved without the loss of signal to noise ratio (SNR). The performance of our approach at lower field strengths (1.5–3T) will require further study. It should be noted, however, that the relative relaxation rates of plaque components do not change at these fields. The exception to this is intraplaque hemorrhage, which is significantly easier to detect with T1 and UTE MRI at 1.5–3T because the longitudinal relaxivity (r1) of iron drops dramatically at higher fields. All the sequences described in this paper can be performed with advanced cardiorespiratory gating and motion compensation. Translation of our approach on current clinical systems is thus feasible.

Limitations and Conclusion

Although our study is to date the largest (n=28) to investigate human coronaries with multicontrast MRI and co-registered histology, the procurement of a broad range of human coronary artery specimens remains challenging. Certain plaque features such as hemorrhage and thrombus were thus not seen in our dataset. If present, however, the iron products in blood and fresh thrombus should be detected with our protocol by producing hyperintensity on the T1 and UTE images. The harvested coronary arteries were imaged with several modalities, and the nature of our study thus required tissue fixation to be performed. This however has been shown not to affect plaque assessment by MRI compared with unfixed specimens (30).

In conclusion, we use a novel combination of MRI contrast (T1, T2 and UTE) to image atherosclerotic plaque in human coronary arteries. We show that plaque classification with this approach compares extremely favorably with histology. The addition of UTE MRI adds significant value by allowing plaque calcification to be accurately detected. Our study further underscores the potential of coronary plaque classification by MRI, and its potential use to define the clinical stage of coronary atherosclerosis and guide patient management.

Figure 1. T1, T2 and UTE images of a calcified lipid-rich necrotic plaque.

L= Lumen. T1-weighting (A) reveals a large plaque with two focal areas of profound hypointensity. These areas are also hypointense on (B) the T2 image but are isointense on (C) the UTE image, consistent with foci of calcification. Profound hypointensity is seen in the plaque on the T2 weighted image (B), consistent with a lipid-rich necrotic core. Pentachrome staining of the plaque cross section (D) correlates extremely well with the MRI. Segmentation of the histological section (E, F) shows the lipid-rich necrotic core (light blue) and the foci of calcification (red dots, white arrows). The plaque was correctly classified as a Type IV–VB plaque by MRI.

Figure 2. A coronary artery plaque with a non-calcified lipid-rich necrotic core.

No profound areas of hypointensity are seen on the T1W image of the plaque (A), exluding plaque calcification. The plaque interior, however, is hypointense on the T2W image (B), consistent with a lipid-rich necrotic core. Plaque histology (D) and segmentation (E, F) correlate well with the MRI. L= lumen, light blue =lipid-rich necrotic core. The plaque was correctly classified as a Type IV–VA lesion by MRI.

Figure 3. Example of plaque misclassification by MRI.

Profound signal hypointensity (arrow) is seen on the T1W image (A) with corresponding recovery of signal in the UTE sequence (C). A corresponding small localized area of hypointensity is seen on the T2W image (B), which could be due either to an isolated focus of calcification or due to calcification of a small lipid-rich necrotic core. (D–F) Histology reveals that the hypointense focus is produced by diffuse calcification (red dots) of a small lipid-rich necrotic core (light blue). The plaque was incorrectly classified by MRI as VIIIB (fibrocalcific without a lipid-rich necrotic core). Only 2/20 lipid-rich necrotic cores were missed by MRI in the study, both of which were small and focal (see panels D, E).

Figure 4. T1W and UTE multiplanar reformats (MPRs) of the left anterior descending coronary artery.

T1-weighted (A) and UTE (B) MPRs of the LAD show two large plaques on opposite sides of the vessel wall. Cross-sectional images through these plaques are shown with T1W (C, E) and UTE (D, F). In both plaques, but in particular in the lower one (A, B, E, F), areas of profound signal hypointensity are seen on T1W images but not on the UTE images, consistent with foci of plaque calcification.

Abbreviations

AHA

American Heart Association

CAD

coronary artery disease

LAD

left anterior descending artery

MPR

multiplanar reformat

MRI

magnetic resonance imaging

UTE

ultra-short echo time

T1

T1-weighting

T2

T2-weighting

R2*

transverse relaxation rate