The added value of longitudinal black-blood cardiovascular magnetic resonance angiography in the cross sectional identification of carotid atherosclerotic ulceration

Stroke is the third most common cause of death and a leading cause of serious, long-term disability worldwide [1]. Extracranial carotid disease accounts for 20–30% of such ischemic events [2]. Carotid atherosclerotic ulceration is a significant contributing source of cerebral ischemic incidents or stroke, along with other plaque features such as intraplaque hemorrhage and thrombosis [3‐10].

Although angiography, specifically digital subtraction angiography (DSA), time-of-flight cardiovascular magnetic resonance (CMR) angiography, or computed tomography angiography (CTA), has become the standard approach to carotid artery disease assessment, these modalities are limited by their inability to show the outer wall of the vessel. Black-blood CMR can depict the actual plaque size by imaging both the lumen and the outer wall of an artery. Studies comparing in vivo black-blood CMR with corresponding endarterectomy specimens confirm the ability of black-blood CMR to accurately visualize atheroma size [11‐13]. On the other hand, time-of-flight CMR angiography yields lower estimates of luminal narrowing compared to black-blood CMR angiography [13, 14], and underestimates the presence of high-risk plaque features [15]. These findings highlight the need for an improved clinical vessel wall imaging method to accurately grade the severity and stage of carotid atherosclerosis [15].

Previous studies have shown good sensitivity and specificity in the identification of carotid atherosclerotic plaque components and the condition of the fibrous cap overlying the necrotic core using multisequence high-spatial-resolution cross-sectional CMR imaging [11, 16‐20]. Identification of fibrous cap rupture is highly associated with recent transient ischemic attack or stroke [17]. Furthermore, studies have demonstrated that detection of carotid atherosclerotic ulceration is possible using these multisequence cross-sectional CMR images [21, 22]. However, cross-sectional CMR images may suffer from partial volume effects due to the intrinsic morphological changes of the carotid bulb [23‐25]. Flow artifacts that mimic plaques [23] and may potentially confound the clear detection of ulcerations occur frequently in this region. The addition of longitudinal black-blood (BB) CMR angiography may reduce errors in classification due to partial volume effects by producing images with greater resolution in the long-axis view of the carotid artery. Longitudinal BB CMR angiography is therefore advocated for pre-operative assessment of the long-axis extent of advanced lesions and for determining the precise location of the carotid bifurcation [26].

In this study, we hypothesize that adding the longitudinal BB CMR angiography to multisequence high-spatial-resolution cross-sectional CMR images can improve the ability of CMR in the characterization of carotid atherosclerotic ulceration.

Of the 32 plaques examined by histology, 17 contained a total of 21 ulcers (14 plaques had a single ulcer, 2 had two ulcers, and 1 had three ulcers). After matching with the CMR images, of the 21 ulcers, 8 occurred on only one cross-sectional CMR slice and 9 occurred on three or more slices.

During round 1 (cross sectional review only), 5 true ulcers were detected. Another six ulcers were detected by adding the BB CMR angiography in round 2. The ulcers detected were significantly larger by histology (median 20 mm³ vs. 2.8 mm³, p = 0.034).

The Spearman's rank correlation between the volume measurements of the ulcers detected by both CMR and histology in round 2 was 0.60 (n = 11, p = 0.053) (Figure 4).

The sensitivity from rounds 1 and 2 were graphed versus ulcer volume by histology, where at each point on the graph, only ulcers of that size or greater were included (Figure 5). In round 2, the sensitivity was 52.4% for all ulcers (n = 21) and 80.0% for ulcers at least 6 mm³ (n = 10). By contrast, the sensitivity from round 1 did not show a clear trend versus size. It was 23.8% and 30.0% for all ulcers and ulcers at least 6 mm³, respectively.

Due to an additional false positive, the specificity decreased slightly from 88.2% in round 1 to 82.3% in round 2. However, both the positive and negative predictive values increased modestly in round 2 (from 71.4% to 78.6% and from 48.4% to 58.3%, respectively). When considering only ulcers at least 6 mm³ by histology, the negative predictive value improved from 73.1% to 90.0%.

In round 2, the true positive ulcers by CMR had larger volumes than the false positives (median 22 mm³ vs. 4.9 mm³, p = 0.005) (Figure 6). In particular, the three false positives were all less than 7 mm³ by CMR whereas the true positives were all greater. 100% specificity could have been achieved without sacrificing sensitivity simply by discarding those CMR measurements. A similar result, though not significant (p = 0.19), was seen in round 1.

In this study, the improvement of ulcer identification can be attributed to the combination of cross-sectional imaging, which has greater resolution in the axial plane but two millimeter slice thickness, and longitudinal BB CMR angiography, which provides higher resolution images along the long-axis in the longitudinal direction. This combination improves the ability to identify carotid plaque ulceration by reducing errors in interpretation caused by mis-registration due to partial volume effects (Figure 3) [23‐25]. On the other hand, the addition of the longitudinal BB CMR angiography did result in increased false positives. However, since both the positive and negative predictive values increased modestly, higher sensitivity represents improvement by the addition of the longitudinal BB CMR in the ulcer identification.

We also demonstrate that the sensitivity of ulcer detection depends on the size of ulcers. When longitudinal BB CMR angiography was added to the cross-sectional CMR images, the sensitivity was 80% when only ulcers with volume at least 6 mm³ were evaluated compared to a sensitivity of 52.4% for all ulcers. There were no ulcers between 3.5 and 6 mm³ (Figure 5) and the overall sample size was small, so the threshold of 6 mm³ could not be estimated precisely. We believe that the dependence of sensitivity on ulcer size is related to the spatial resolution provided by the cross-section CMR images and longitudinal BB CMR angiography. As 3.5–6 mm³ corresponds to an average of 2.4–2.9 pixels in each of the three dimensions of cross-sectional CMR images and BB CMR angiography (0.625 mm/pixel with 2 mm thickness for each 2D image), our results suggest the minimum requirement of spatial resolution of CMR for consistent ulcer identification is at least 2–3 pixels.

Low sensitivity occurred in the first round of review with cross-sectional CMR images alone. This can be attributed to two factors. First, presence of partial volume effect and/or flow artifact in the cross-sectional images only led to more false negatives, which subsequently caused lower sensitivity (Figure 3). Second, a conservative approach was used for identifying ulcers by CMR, i.e. ulcers were only identified when the reviewers were very confident of the assessment.

Drawing direct comparisons between the sensitivity and specificity in this study with others is difficult because different imaging modalities and different postprocessing methods were used for ulcer detection [9, 31‐34]. For ulcer detection, conventional angiography generated a sensitivity of 46% and a specificity of 74%, as it used limited views (e.g., anterior-posterior/oblique/lateral acquisitions) in the NASCET trial [9]. Single-slice CTA resulted in a sensitivity of 60% and a specificity of 74% [31]. Multi-detector CTA showed a sensitivity of 94% and a specificity of 99% [32]. A combination of axial images, maximum intensity projection, multi-planar reformatting, and volume rendering on Multi-detector CTA resulted in higher sensitivity (94%) and specificity (99%) [34]. Therefore, influence of imaging techniques and post processing methods should be taken into account when comparing the accuracy of ulcer detection.

Furthermore, direct comparison between studies of ulceration is difficult because definitions of ulceration vary both among imaging modalities and in histology [9, 31‐33]. In imaging modalities, ulceration is defined by its shape (crater penetrating into a stenotic plaque or double density on ''en face'' view), its relationship to the vessel lumen (extension of contrast beyond the vascular lumen into surrounding plaque), or irregularity of luminal surface (break in the surface of the plaque with a depth of ≥1 mm) [9, 31‐33]. In histology, ulceration is defined by the integrity of the plaque surface (disruption in the lining of the plaque or a pit and depression in the plaque, fissuring of the fibrous cap, or break in the surface of the plaque), and/or by penetration depth (approximately ≥1 mm or ≥2 mm) [9, 31‐33].

In this study, 53% (17 of 32) plaques examined by histology had ulcers. This is consistent with the prevalence of ulceration reported in other studies [35, 36]. Park et al reported that 77% of carotid lesions removed from symptomatic patients (transient ischemic attack, amaurosis fugax, prior stroke) and 60% of plaques removed from asymptomatic patients with high-grade stenosis had ulceration [35]. Saam et al reported 25 of 46 (54%) carotid plaques had fibrous cap rupture [36]. All these patients have similar demographic characteristics as our patients in this study. Therefore, the prevalence of ulceration reported in this study is representative of typical demographic characteristics in a population of subjects who had experienced a transient ischemic attack, amaurosis fugax, or stroke within 6 months of their surgery and had >50% carotid stenosis or who were asymptomatic with >80% carotid stenosis by duplex ultrasound.

Our study had several limitations. Longitudinal BB CMR angiography is acquired by a 2D multi-slice double inversion recovery sequence through the centers of internal and external carotid arteries at the carotid bifurcation. If the artery is tortuous, not all segments of the carotid can be covered by this sequence, therefore, ulceration may not be well identified. Implementation of 3D isotropic resolution carotid wall CMR may improve visualization of all segments of carotid artery and allow for better identification of ulceration [37]. In addition, BB CMR angiography may still be insufficient for cases where flow artifact confounds ulcer identification. Use of a motion sensitized driven equilibrium (MSDE) turbo spin echo sequence on 3T CMR may assist in suppression of plaque-mimicking flow artifacts and improve image quality for high-resolution black-blood imaging of carotid arteries [38, 39]. This sequence has demonstrated promise in the identification of carotid atherosclerotic ulceration [38]. Furthermore, use of 8-element phased array carotid coils will improve over 4-element coils in the longitudinal coverage and vessel wall signal-to-noise ratio and contrast-to-noise ratio [40]. The number of patients examined is relatively small. Due to the limited number of subjects included, accuracy of adding longitudinal BB CMR angiography needs to be further confirmed by including larger number of subjects in a future study. In addition, due to the small number of subjects, we did not correlate the presence and sizes of ulceration with neurological events in this study. Finally, the study was not blinded.

This study demonstrates that adding longitudinal black-blood CMR Angiography to multisequence high-spatial-resolution cross-sectional CMR images improves identification of carotid atherosclerotic ulcerations.