Quantitative evaluation of carotid atherosclerotic vulnerable plaques using in vivo T1 mapping cardiovascular magnetic resonaonce: validation by histology

It is well evidenced that cerebrovascular vulnerable atherosclerotic plaque is the key etiology of ischemic stroke [1]. Histologically, intraplaque hemorrhage (IPH) and large necrotic core (NC) have been considered as the key features of vulnerable plaques [2]. In addition, investigators found that stage of IPH was associated with cardiovascular events [3]. Therefore, it is important to accurately characterize vulnerable plaque components.

Multi-contrast cardiovascular magnetic resonance (CMR) vessel wall imaging, including time-of-flight (TOF), T1-weighted, and T2-weighted sequences, has become an ideal non-invasive technique to assess carotid plaque components [4]. In multi-contrast CMR imaging techniques, each sequence plays a specific role in the identification of specific plaque component. Pre- and post-contrast T1-weighted imaging has been largely used to evaluate NC with good agreement with histology [5]. The sequence of magnetization prepared rapid gradient echo acquisition (MP-RAGE) has been demonstrated to be the best technique in characterizing IPH among three different T1-weighted sequences [6]. It is evidenced that the stage of carotid artery IPH can be distinguished by combining TOF, T1-weighted, and T2-weighted images [7].

Since plaque components are determined according to their signal intensity on different contrast images and CMR signal is mainly dependent on the values of T1 and T2 relaxation, it is possible to distinguish plaque components by quantitative imaging. Recently, quantitative CMR imaging techniques, such as T1 mapping, have been proposed for plaque characterization [8‐11]. Although investigators utilized ex-vivo T1 mapping to assess intracranial plaques [8], few studies investigated the in-vivo T1 values of carotid plaque components. Ota and colleagues measured the in-vivo longitudinal relaxation rates (R1 = 1/T1) of carotid homogenous plaque components which could not be applied to characterization of those plaques with inhomogeneous and complicated components [9]. Subsequently, Qi et al. developed a 3D SNAP with golden angle radial k-space sampling (GOAL-SNAP) sequence for measuring T1 values of vessel wall and this technique showed a potential in identifying IPH and monitoring its progression [10]. However, the capability of GOAL-SNAP sequence in characterizing vulnerable plaques has not been validated by histology.

Here, we hypothesized that different plaque components have different T1 values and the T1 values of vessel wall can be used to identify vulnerable plaque features. This study sought to validate the usefulness of in vivo T1 mapping in assessing carotid vulnerable plaque compositional features by histology.

Of 34 recruited patients, 15 (44.1%) had hypertension, 3 (8.8%) had diabetes, 2 (5.9%) had hyperlipidemia, and 22 (64.7%) had a history of cardiovascular disease. Of 34 subjects, 1 failed to receive CEA surgery due to perioperative cardiac event and 4 had no intact specimens because of operative procedure. In total, 29 qualified specimens were collected for analysis. The clinical information is detailed in Table 2.

In this study, we calculated the T1 values of carotid atherosclerotic plaque components which were identified on multi-contrast vessel wall imaging or validated by histology. We found that T1 values of IPH, NC and LM were significantly different which suggests the potential capability of T1 mapping in the identification and segmentation of carotid plaque components. Moderate to good agreement was found in the identification, classification and quantitative measurements of IPHs between T1 mapping and histology.

In this study, significant difference in T1 values was found among NC, IPH and LM validated by multi-contrast CMR images or histological specimens. Few previous studies investigated the in-vivo T1 values of carotid plaque components due to the challenges of high-resolution vessel wall T1 mapping. Previously, ex-vivo intracranial plaques from 20 cadavers were studied in which the T1 values of NC, fibrous tissue, fibrous cap, wall and calcification were reported [8]. However, the ex-vivo plaques have different conditions, such as temperature and exchanges of molecules, compared with in-vivo ones. Recently, Ota et al. measured the longitudinal relaxation rates (R1 = 1/T1) of in-vivo carotid plaque components on pixel basis validated by histology [9]. The R1 of NC, IPH and fibrous tissue was reported to be 0.4–1.2 s^− 1 (T1: 833–2500 ms), ≥ 1.5 s^− 1 (T1 ≤ 667 ms), and 0.2–0.8 s^− 1 (T1: 1250–5000 ms), respectively. In contrast, other in-vivo studies at 3 T reported that the T1 value for the vessel wall, muscle, and IPH was 1195 to 1213 ms, 1027 to 1219 ms, and 371 ± 93 ms (the minimum value), respectively [10, 18, 19]. The T1 value of NC in the present study was comparable with previous studies. In addition, the T1 value of IPHs which were identified on multi-contrast CMR images and histology was 907 ± 354 ms and 651 ± 253 ms, respectively. This difference might be caused by a larger proportion of old IPH which showed a longer T1 value than fresh IPH in multi-contrast imaging samples than that in histological specimens (66.1%). The difference might be contributed by the mismatched resolution between CMR imaging and histology. The IPHs identified on multi-contrast images might include part of NC which would cause a longer T1 value. The reported T1 values of LM or NC identified by multi-contrast imaging and histology were similar. Additional value of the present study was that the in-vivo T1 value of LM on 3 T validated by histology was firstly reported. The differentiable T1 values of carotid plaque components suggest the potential of automatic segmentation of plaque components using T1 mapping.

T1 mapping was found to have a moderate agreement with histology in identification and classification of carotid IPHs. In the present study, the IPHs were divided into two types according to the percentage of fresh hemorrhage on histology. We found that Type 1 IPH showed significantly shorter T1 value than Type 2 IPH and T1 values had strong correlation with the proportion of area of fresh hemorrhage occupying the total area of IPH, indicating that lower T1 values signify fresher hemorrhage which represents a higher risk of carotid plaque. Histologically, fresh IPH was mainly consisted of intact red blood cells, polymorphonuclear cells, lymphocytes, and scattered macrophages [7]. The intact red blood cell with intracellular methemoglobin in fresh IPH might shorten the T1 values. In addition, a threshold segmentation was also utilized to quantify carotid IPH and good agreement was found in measuring the area of IPHs between T1 mapping and histology in the present study. The threshold segmentation is appealing because it can be used across different centers. In the present study, T1 mapping images had a low accuracy and specificity in identification of NC and LM which might be caused by the overlaps of their T1 values. The NC has complex components including cholesterol crystal, debris of apoptotic cells and particles of calcium which are attributed to wide range of T1 value. LM shows a long T1 value which might be underestimated by the GOAL-SNAP imaging due to the intrinsic limitation of this technique for the long T1 value validated by spin-echo sequence [10]. Therefore, it was challenging to accurately identify and segment the NC and LM using T1 mapping. In addition, the identification and segmentation of plaque components using T1 mapping were based on the boundaries of arterial wall which dominantly determined by T1-QIR images of multi-contrast vessel wall imaging. Because the wall boundaries on T1 mapping images were not clear, the T1 mapping images need to be combined with at least one black blood T1 weighted sequence to identify plaque components.

There are several limitations in this study. First, the registration of CMR images and histology might be influenced by the mismatched resolution. The histological sections were excised every 0.5 mm with 10 μm thickness. In contrast, the thickness of CMR images was 2 mm. High-resolution CMR vessel wall imaging along with decreased partial volume effects is warranted in future studies. Second, fibrous cap, as a key feature of vulnerable plaques, was not evaluated due to the limited spatial resolution. A previous histological study showed that the minimum thickness of thin fibrous cap was 200 μm [20], suggesting that a higher resolution CMR imaging will be required for assessment of fibrous cap. Third, calcification was not evaluated in this study due to the low signal-to-noise ratio of calcification and reliable T1 cannot be fitted. Qi et al. set the value of T1 and T2 of calcification as zero to realize the segmentation of calcification [18]. Fourth, fibrous tissue was not evaluated in this study. Masson Trichrome stain for the assessment of fibrous tissue is warranted in the future. Fifth, the sample size of our study was relatively small and most of results were from slice-level analysis which might cause over/under representation of some subjects and introduce bias. A larger sample size with the consideration of inter-subject variability was warranted for the validation of T1 mapping in identifying plaque components with enough training and test datasets. Finally, histological sections were ex-vivo samples in which plaque components might have morphological changes to some extent. Gamble et al. reported that about 7.8% of shrinkage was determined in whole ex-vivo specimens [17]. However, the shrinkage of each plaque component has not been investigated. Furthermore, the ex-vivo specimens might have different shape compared with the in-vivo sections.

CMR T1 values of carotid plaque components, particularly for IPH, are differentiable, and the stage of IPH can be classified according to T1 values, suggesting the potential capability of assessment of vulnerable plaque components by T1 mapping.