High resolution 7T MR imaging in characterizing culprit intracranial atherosclerotic plaques

Abstract

Background

Current imaging modalities underestimate the severity of intracranial atherosclerotic disease (ICAD). High resolution vessel wall imaging (HR-VWI) MRI is a powerful tool in characterizing plaques. We aim to show that HR-VWI MRI is more accurate at detecting and characterizing intracranial plaques compared to digital subtraction angiography (DSA), time-of-flight (TOF) MRA, and computed tomography angiogram (CTA).

Methods

Patients with symptomatic ICAD prospectively underwent 7T HR-VWI. We calculated: degree of stenosis, plaque burden (PB), and remodeling index (RI). The sensitivity of detecting a culprit plaque for each modality as well as the correlations between different variables were analyzed. Interobserver agreement on the determination of a culprit plaque on every imaging modality was evaluated.

Results

A total of 44 patients underwent HR-VWI. Thirty-four patients had CTA, 18 TOF-MRA, and 18 DSA. The sensitivity of plaque detection was 88% for DSA, 78% for TOF-MRA, and 76% for CTA. There's significant positive correlation between PB and degree of stenosis on HR-VWI MRI (p < 0.001), but not between PB and degree of stenosis in DSA (p = 0.168), TOF-MRA (p = 0.144), and CTA (p = 0.253). RI had a significant negative correlation with degree of stenosis on HR-VWI MRI (p = 0.003), but not on DSA (p = 0.783), TOF-MRA (p = 0.405), or CTA (p = 0.751). The best inter-rater agreement for culprit plaque detection was with HR-VWI (p = 0.001).

Conclusions

The degree of stenosis measured by intra-luminal techniques does not fully reflect the true extent of ICAD. HR-VWI is a more accurate tool in characterizing atherosclerotic plaques and may be the default imaging modality in clinical practice.

Introduction

Intracranial atherosclerotic disease (ICAD) is one of the leading causes of ischemic stroke in the world accounting for 10% in the United States and up to 50% in Asia. ¹ The diagnostic tools currently used for the evaluation of patients with stroke might not be adequate to detect the presence and severity of ICAD. We have recently reported that ICAD was underestimated in stroke patients previously diagnosed as having cryptogenic stroke. ² Traditional imaging methods, such as computed tomographic angiography (CTA), time of flight magnetic resonance angiography (TOF-MRA), and digital subtraction angiography (DSA), are generally adequate at assessing vascular luminal stenosis. ³ However, these modalities are suboptimal at detecting the presence of non-stenotic plaques, with outward remodeling and large plaque burden (PB). ⁴ PB is usually defined as the percentage of plaque area out of the total vessel area at the maximal stenosis lesion site, within a cross sectional plane. High resolution-vessel wall imaging (HR-VWI) allows for a better detection and characterization of PB, which is a predictor of both cardiovascular and cerebrovascular events and death.^5,6 Treatment failure and stroke recurrence may be related to the poor characterization of PB in patients with non-stenotic plaques. ⁷ This study is aimed at comparing different imaging modalities (HR-VWI MRI, CTA, TOF-MRA, and DSA) in the identification and characterization of intracranial atherosclerotic culprit plaques.

Methods

Patient population and data collection

This study was approved by our institutional review board. Patients with ischemic stroke/transient ischemic attack (TIA) attributed to ICAD were prospectively enrolled. ICAD stroke etiology was adjudicated based on TOAST and ASCOD criteria.^8,9 The potential cause for atherothrombotic stroke is defined by ASCOD as: ipsilateral atherosclerotic stenosis between 50 and 99% in an intracranial or extracranial artery supplying the ischemic field. ⁹ Based on this criterion, the most stenotic plaque in the territory of a stroke or TIA was deemed as the culprit plaque in case of multiple plaques in the same vascular territory.^1,2 Diffusion weighted imaging (DWI) lesions were identified on MRI to determine the presence of strokes. In case of TIAs, culprit plaque was determined in the ipsilateral vascular territory where symptoms localized.

Patients were excluded if they had any contraindication for 7T MRI, were medically unstable, were agitated/claustrophobic, had a body mass index ≥ 30, or had a glomerular filtration rate <60 mL/min per 1.73 m². Written consent was obtained from all subjects prior to enrollment. Conventional imaging modalities performed as part of stroke work up were analyzed: CTA, TOF-MRA, and DSA. Additionally, DWI was analyzed to measure the infarct volume. Demographic and clinical information was acquired from electronic medical records.

Imaging acquisition and identification of culprit plaque

HR-VWI MRI images were acquired on a GE MR950 7T scanner, using an 8-channel head coil (GE Healthcare, Waukesha, WI). The following 3D isotropic sequences were acquired: T1-weighted fast-spin-echo (CUBE), T2-weighted CUBE, and susceptibility-weighted angiogram. TOF MRA was obtained on a 3T scanner. Technical parameters for acquisition are described in the Supplementary Table 1. Gadobutrol (Gadavist, Bayer Pharmaceuticals, New Jersey, USA) was administered intravenously (0.1 mmol/kg), and a postcontrast 3D TI CUBE sequence was obtained after ≈5 min. Images were analyzed with the Picture Archiving Communication System (PACS, Carestream Vue, version 12.1.6.1005). The ipsilateral intracranial vasculature of the symptomatic territory was analyzed in multiple planes with T1-weighted pre- and post-contrast images to identify the presence of culprit atherosclerotic plaques. Lumen patency was assessed at the site of wall thickening identified on MR images and compared with the corresponding location on TOF-MRA, CTA, and DSA.

Measurements and definition of plaque characteristics

A plaque was defined as wall thickening on the cross-sectional view of a segment using its adjacent proximal, distal, or contralateral vessel segment as a reference. ¹⁰ A region of interest (ROI) was carefully drawn to determine the following parameters: the area of the vessel lumen at maximal stenosis, the area of the outer vessel at maximal stenosis, and the area of the vessel lumen proximal or distal to maximal stenosis. To measure the degree of stenosis, the lumen at maximal stenosis is compared with the lumen of the parent artery proximal to the lesion (Figure 1). Cross-sectional orthogonal views were used to characterize PB. Multiplanar views were used to perform measurements of the vessel's wall and lumen using PACS. The best maximal intensity projection was used to perform all the plaque measurements. For each culprit plaque segment, diameters and areas of interest were measured and calculated as follows:

• Degree of stenosis based on the “Warfarin-Aspirin Symptomatic Intracranial Disease” (WASID) trial: ¹¹

  $$\left({\displaystyle \frac{\mathrm{Diameter}\mathrm{of}\mathrm{lumen}\mathrm{at}\mathrm{proximal}\mathrm{normal}\mathrm{segment}-\mathrm{diameter}\mathrm{of}\mathrm{lumen}\mathrm{at}\mathrm{maximal}\mathrm{stenosis}}{\mathrm{Diameter}\mathrm{of}\mathrm{lumen}\mathrm{at}\mathrm{proximal}\mathrm{normal}\mathrm{segment}}}\right)*100$$
• Degree of stenosis based on ROI area:

  $$\left({\displaystyle \frac{\mathrm{Area}\mathrm{of}\mathrm{lumen}\mathrm{at}\mathrm{proximal}\mathrm{normal}\mathrm{segment}-\mathrm{area}\mathrm{of}\mathrm{lumen}\mathrm{at}\mathrm{maximal}\mathrm{stenosis}}{\mathrm{Area}\mathrm{of}\mathrm{lumen}\mathrm{at}\mathrm{proximal}\mathrm{normal}\mathrm{segment}}}\right)*100$$
• Plaque Burden (PB):

  $$\left({\displaystyle \frac{\mathrm{Area}\mathrm{of}\mathrm{outer}\mathrm{vessel}\mathrm{at}\mathrm{maximal}\mathrm{stenosis}-\mathrm{Area}\mathrm{of}\mathrm{lumen}\mathrm{at}\mathrm{maximal}\mathrm{stenosis}}{\mathrm{Area}\mathrm{of}\mathrm{outer}\mathrm{vessel}\mathrm{at}\mathrm{maximal}\mathrm{stenosis}}}\right)*100$$
• Remodeling Index (RI):

  $$\left({\displaystyle \frac{\mathrm{Area}\mathrm{of}\mathrm{vessel}\mathrm{at}\mathrm{maximal}\mathrm{stenosis}}{\mathrm{Area}\mathrm{of}\mathrm{vessel}\mathrm{at}\mathrm{proximal}\mathrm{normal}\mathrm{segment}}}\right)*100$$

Figure 1.

Plaque analysis based on 2D diameters measurements and ROIs. This figure is copyrighted to Edgar Samaniego and published with permission.

The degree of stenosis for each culprit plaque (WASID criteria) was calculated on all the four modalities: HR-VWI MRI, CTA, TOF-MRA, and DSA. Additionally, ROI degree of stenosis was measured in 7T HR-VWI. PB and RI were calculated using HR-VWI based on 2D diameters to determine the areas used on these assessments (area =  $\pi .\left(\frac{\mathrm{diameter}}{2}\right)$ ²). The areas obtained with 2D measurements were compared with the alternative way of measuring areas, which includes manual ROI selection. The pattern of arterial remodeling at the plaque was categorized as positive remodeling if the RI > 1.05.

Infarct volume measurements

The MRIs obtained after stroke/TIA were reviewed to characterize DWI lesions. The ROI selection tool in PACS was used to measure the volume of stroke in each patient.

Inter-rater culprit plaque detection

Culprit plaque identification was compared between two separate adjudicators. Both reviewers are experienced neuroradiologists trained in plaque detection on high resolution imaging. Before adjudication, each rater reviewed a slide tutorial demonstrating different plaque morphologies on 7T-MRI. Each reviewer analyzed every image modality and filled a questionnaire (Supplementary Material). Reviewers determined the presence and location of plaques after being provided with the location of the stroke/TIA. In case of disagreement, a senior neuroradiologist and a senior neurologist who had access to all the patient information were the final adjudicators of culprit plaques by consensus. This final adjudication was used to determine the accuracy of plaque detection and localization by the two raters, and to calculate the interrater agreement.

Statistical analysis

Continuous variables are presented as mean (standard deviation [SD]), and categorical variables are presented as frequency and percentages. Values measured were tested for normality by using the Shapiro-Wilk and Kolmogorov-Smirnov tests. For normally distributed data, continuous data were compared by using Students T-Test and categorical data was analyzed by using Pearson chi-squared test. For non-normal variables, we utilized the Wilcoxon signed rank test for comparison of medians. Agreement of quantitative measurement was assessed with Bland-Altman plots. All correlations between different variables including PB, RI, and degree of stenosis on different modalities were tested using Pearson's correlation test. The correlation of PB and infarct volume was evaluated using Spearman's Rho. One-way ANOVA was performed to calculate differences of time between the acquisition of HR-VWI and the other imaging modalities. Linear regression analysis was used to determine the association of degree of stenosis to PB and RI. Interobserver agreement on the determination of the culprit plaque for every imaging modality was evaluated with Cohen κ. A 2-sided p < 0.05 was considered significant. All statistical analyses were performed with SPSS Statistics 25.0 (IBM, Armonk, New York, USA).

Results

Study subjects

Fifty-four patients with symptomatic ICAD who underwent HR-VWI MRI were screened. A total of 44 patients with culprit plaques were included in the final analysis (Supplementary Figure 1). Thirty-four patients had a CTA, 18 had TOF-MRA, and 18 had a DSA. The primary treating physician determined which images were required for diagnostic purposes. As part of the study, every patient underwent HR-VWI. The order of image acquisition for each patient is depicted in Supplementary Table 2. The mean age was 60.9 ± 13.8 years, and 56.8% were women. The mean time (months ± SD) from stroke/TIA onset to imaging was: HR-VWI MRI = 5.1 ± 9.6, CTA = 1.5 ± 3.9, TOF-MRA = 1.9 ± 4.2, and DSA = 3.9 ± 6.5 (p = 0.04) (Table 1). The mean time (months ± SD) between HR-VWI and the acquisition of other imaging modalities was: CTA = 3.5 ± 6.3, TOF-MRA = 8.4 ± 11.6, and DSA = 2.6 ± 5.6 (p = 0.06).

Table 1.

Baseline characteristics.

Variable	Values
Patients (n = 44)
Age (mean ± SD, in years)	60.9 ± 13.8
Women (%)	25 (57)
Hypertension (%)	31 (71)
Diabetes (%)	9 (21)
Smoking (%)	18 (41)
Territory of infarction
Frontal (%)	8 (18)
Temporoparietal (%)	9 (20)
Basal ganglia/thalamic (%)	8 (18)
Occipital (%)	6 (14)
Cerebellar (%)	4 (10)
Midbrain – pontomedullary (%)	9 (20)
Anterior circulation (%)	25 (57)
Posterior circulation (%)	19 (43)
Culprit plaque location
ICA (%)	7 (16)
MCA (%)	17 (39)
PCA (%)	9 (20)
BA (%)	8 (18)
VA (%)	3 (7)
Degree of stenosis (mean ± SD) using WASID formula
HR-VWI MRI (WASID)	47 ± 21%
HR-VWI MRI (ROI)	58 ± 23%
DSA	54 ± 26%
TOF-MRA	44 ± 31%
CTA	41 ± 28%
Time from Stroke to Imaging (mean ± SD, in months)
HR-VWI MRI	5.1 ± 9.6
DSA	3.9 ± 6.5
TOF- MRA	1.9 ± 4.2
CTA	1.5 ± 3.9

BA = basilar artery; ICA = internal carotid artery; MCA = middle cerebral artery; PCA = posterior cerebral artery; VA = vertebral artery; TI = transient ischemic attack; HR-VWI MRI = high resolution vessel wall imaging MRI; DSA = digital subtraction angiogram; CTA/MRA = CT and MR angiogram.

WASID formula: “Comparison of Warfarin and Aspirin for Symptomatic Intracranial Arterial Stenosis” formula for calculating the degree of stenosis = [(diameter of lumen at proximal normal segment - diameter of lumen at maximal stenosis)/(diameter of lumen at proximal normal segment)]*100

Data shown as n (percentage) of patients unless otherwise specified.

Identification of plaques and degree of stenosis

When compared to 7T HR-VWI MRI, the sensitivity for culprit plaque detection was 76% (26/34) on CTA, 78% (14/18) on TOF-MRA and 88% (16/18) on DSA (Figure 2). For the purpose of this analysis, HR-VWI was considered the gold standard in detecting culprit plaques. HR-VWI characterizes degree of stenosis, PB and RI, therefore, all the other imaging modalities were compared to HR-VWI.

Figure 2.

Culprit plaque identification for every image modality. High-resolution vessel wall imaging (HR-VWI) is considered the gold standard.

The mean ± SD diameter of the lumen at maximal stenosis was 1.33 ± 0.69 mm on HR-VWI MRI, 1.27 ± 0.62 on CTA, 1.16 ± 0.63 mm on TOF-MRA and 1.09 ± 0.72 mm on DSA. The mean ± SD measured ROI area of the lumen at maximal stenosis was 2.39 ± 2.02 mm², and the mean ± SD measured ROI area of the outer vessel perimeter at maximal stenosis was 15.46 ± 6.79 mm² (Figure 3).

Figure 3.

A patient with a history of vertebro-basilar insufficiency symptoms presented with a cerebellar infarct. The measurement of the vertebral artery (VA) culprit plaque on all modalities is illustrated: upper row shows measurement of the stenotic segment and lower row of the normal proximal parent artery segment. A ROI is delineated at the level of the plaque (A, upper row) and the normal vessel (A, lower row) in orthogonal planes. The degree of stenosis (DS) is determined by ROIs = 78.1% (A) and diameter measurements = 68.6% (B). DS vary significantly (78 to 40%) based on the imaging modality used: DSA (C) = 67.7%, TOF MRA (D) = 54.3%, and CTA (E) = 40%. In this example the plaque has circumferential positive remodeling which directly translates into luminal stenosis. In the DSA (C), the plaque appears to occupy only one side of the VA, but on HR-VWI the plaque extends along the entire arterial segment.

The degree of stenosis based on WASID was not significantly different between different imaging modalities: 47 ± 22% on HR-VWI MRI, 41 ± 28% on CTA, 44 ± 30% on TOF-MRA and 54 ± 26% on DSA (Supplementary Figure 2). However, when determining the degree of stenosis by area ROI analysis on HR-VWI MRI (Figure 3A), significant differences with the conventional WASID measurements were encountered: 58 ± 23% (ROI) versus of 47 ± 22% (WASID) (p < 0.001).

Plaque burden and remodeling index

PB was significantly different when measured by the area of ROIs (mean 87 ± 11 SD) compared to just diameter measurements (mean 92 ± 10 SD) on the Wilcoxon signed ranked test (p < 0.001). The degree of stenosis was positively correlated with PB measured by ROIs (r = 0.60, p < 0.001) and WASID (r = 0.50, p < 0.001) criteria on 7T-VWI MRI. However, PB was not correlated with the degree of stenosis based on WASID criteria on CTA (p = 0.253), MRA (p = 0.144), and DSA (p = 0.168).

The mean RI was 3.1 ± 1.4 based on ROI area, and 3.2 ± 1.5 based on measured diameters (p = 0.149). Forty-one out of 44 plaques (93%) had positive remodeling (RI>1.05). ROI-based measurements showed that higher RIs correlated with less degree of stenosis on HR-VWI MRI (r = −0.43, p = 0.003), but not based on WASID measurements (p = 0.108) on HR-VWI MRI (p = 0.108), CTA (p = 0.751), MRA (p = 0.405) and DSA (p = 0.783).

Plaque burden and remodeling index correlate with degree of stenosis

A linear regression model was used to determine if PB and RI are correlated with degree of stenosis on every imaging modality. A bivariable analysis showed that both PB and RI correlate with degree of stenosis on HR-VWI MRI whether measured through ROIs or through WASID criteria. PB and RI correlated with degree of stenosis based on ROIs: beta coefficient 2.06, p < 0.001 and −0.11, p < 0.001; respectively. PB and RI correlation with degree of stenosis based on WASID measurements: beta coefficient 1.98, p < 0.001; and −0.08, p < 0.001; respectively. PB and RI did not correlate with degree of stenosis as measured by WASID on either CTA (PB: p = 0.158; RI: p = 0.364), TOF-MRA (PB: p = 0.053; RI: p = 0.123), or DSA (PB: p = 0.106; RI: p = 0.344).

Plaque burden and infarct volume

PB did not correlate with infarct volume (Rho = 0.21, p = 0.201) (Supplementary Figure 3).

Inter-rater plaque detection

A total of 114 sets of images (44 patients) were analyzed by each rater. Culprit plaques were identified 98% of time on HR-VWI, 56–88% on CTA, 11–78% on TOF-MRA, and 81–83% on DSA (Table 2). The percentage of accurate plaque detection by the raters, after final adjudication by senior neuroradiologist and neurologist, was: 75–80% for HR-VWI, 50–68% for CTA, 33–78% for TOF-MRA, and 72–83% for DSA (Table 2). A Cohen K analysis demonstrated moderate inter-rater agreement for accurate culprit plaque detection on HR-VWI (k = 0.48, p = 0.001), and suboptimal agreement on the other imaging modalities: CTA (k = -0.18, p = 0.271), TOF-MRA (k = -0.13, p = 0.423), and DSA (k = 0.05, p = 0.814).

Table 2.

Detection of culprit plaque for each rater. High resolution-vessel wall imaging is the gold standard and is compared to the other imaging modalities.

	Rater 1			Rater 2			Rater 1 & 2
	N	% plaque detected	% correct plaque detected	N	% plaque detected	% correct plaque detected	Interrater agreement*
HR-VWI	44	98%	80%	44	98%	75%	p = 0.001
DSA	18	81%	72%	18	83%	83%	p = 0.814
MRA	18	78%	78%	18	11%	33%	p = 0.423
CTA	34	88%	68%	34	56%	50%	p = 0.271

HR-VWI = high resolution vessel wall imaging; DSA = digital subtraction angiogram; CTA/MRA = CT/MR angiogram.

*Based on Cohen kappa statistic for interrater agreement for percent correct plaque detected for each modality.

Significant values are highlighted in bold.

Discussion

The identification and characterization of atherosclerotic plaques varies significantly between imaging modalities. The estimation of PB and RI by ROIs on HR-VWI appears to be an accurate method of assessing the extent of atherosclerotic changes in symptomatic patients. Conventional luminal imaging modalities, such as TOF-MRA, CTA, and DSA are not accurate at determining positive remodeling, PB and RI, and underestimate disease progression.

The conventional determination of degree of stenosis based on WASID criteria underestimates PB. Moreover, PB is positively correlated with the degree of stenosis measured on HR-VWI, but not on CTA, TOF-MRA, and DSA. A comprehensive analysis of the entire arterial segment with ROIs (Figure 1 and 3A) demonstrated a broader disease process than simply determining degree of stenosis with WASID 2D-based measurements. The mean PB calculated based on orthogonal views of the diseased arterial segment and by drawing ROIs was 87%, as opposed to 92% when calculated with 2D diameter measurements. An ROI-based analysis of PB appears to be gold-standard in determining the degree of disease of an arterial segment. ROIs are manually sampled in an orthogonal plane to precisely separate plaque from the arterial lumen, and to define vessel wall thickening. This approach is especially useful with plaques that have positive outward remodeling. PB encompasses the entire spectrum of disease: total wall volume, luminal stenosis, and plaque volume. Previous analysis of carotid stenosis measurements using ROI cross-sectional areas has shown to be a more accurate than changes in 2D diameters. ¹²

HR-VWI is a more accurate at determining the burden of disease compared to other imaging methods. Gong et al. compared 3T-HR-VWI with DSA and TOF-MRA in determining luminal stenosis. ¹³ The degree of stenosis was calculated based on WASID criteria and not with cross-sectional ROIs. HR-VWI had an excellent agreement with DSA, whereas TOF-MRA performed poorly when compared to DSA. ¹³ Liu et al. showed that the degree of stenosis of the middle cerebral artery was accurately determined by 3T HR-VWI and not by CTA. ¹⁴ Measurements were assessed by the WASID 2D method. The authors reported a higher correlation between HR-VWI and DSA, than between CTA maximal intensity projection measurements and DSA. 7T HR-VWI is the best imaging modality clinically available for studying vessel wall. The increased MR signal at 7T renders high spatial resolution and high signal-to-noise ratio. ¹⁵ Recent studies have demonstrated that the use of 7T HR-VWI improves the identification of plaque morphology and composition, even when compared to 3T HR-VWI.^16,17 Therefore, we used 7T HR-VWI for detecting and characterizing culprit plaques.

Most of the patients included in this cohort (93%) had outward positive remodeling, with a PB of 87% based on orthogonal ROI measurements. Positive outer wall remodeling is an important component of plaque morphology, frequently seen in non-stenotic atherosclerotic lesions. ¹⁸ Plaques with positive remodeling have the same effect on stroke risk than plaques without significant luminal stenosis.^19,20 Outer wall positive remodeling has also been associated with a higher risk of microemboli detected by transcranial Doppler. ²¹ A meta-analysis by Lee et al. showed that culprit plaques with positive remodeling are strongly associated with stroke events (OR 6.1; 95% CI 3.2–11.9). ²² Qiao et al. showed that positive remodeling occurs more frequently in the posterior circulation and is more often found on culprit plaques (OR 1.70). ²⁰ In the population-based ARIC study, up to 10.8% of identified lesions by 3T HR-VWI were nonstenotic. ²³ Coronary plaque analysis has shown that positively remodeled plaques have higher lipid content, larger lipid cores, thinner cap fibroatheromas, and higher macrophage counts than plaques with negative remodeling.^24,25 The outer growth of the vessel wall in response to atherosclerosis may lead to higher plaque vulnerability and a more unstable phenotype than plaques with negative remodeling. We have determined that the assessment of positive remodeling is more accurate when it is based on orthogonal ROIs sampling. Our mean RI based on ROIs was 3.1. This is higher than previous reports, and probably is explained by the higher sensitivity of 7T MRI in detecting arterial wall thickening.^20,21

Higher PB has been associated with the presence of symptomatic culprit plaques. ²⁶ PB may increase in diseased vessels without a significant change in luminal diameter (Supplementary Figure 4). ²⁷ HR-VWI had a high sensitivity in identifying plaques with high PB and a low degree of stenosis. Approximately 40% of the plaques with less than 50% luminal stenosis on DSA had high PB as determined by HR-VWI (mean PB = 80%) (Supplementary Figure 5). Other studies have also suggested that up to 50% of high-risk plaques may have a high PB without significant arterial stenosis (<50%). ⁷ Ran et al. showed that every 10% increase in PB would lead to a 2.2-fold higher risk of stroke recurrence. ⁶ PB was positively correlated with the degree of stenosis measured on HR-VWI but did not correlate with the degree of stenosis on CTA, TOF-MRA, and DSA. PB provides both the degree of lumen narrowing and the degree of arterial positive remodeling, and therefore is a more accurate estimation of stroke risk than degree of stenosis alone. ⁷ PB did not correlate with infarct volume, however, the size of infarct volume is determined by many factors, such as collateral flow, use of antiplatelets, blood pressure, among others. ²⁸

Although DSA has been considered the gold standard for quantifying stenosis, non-invasive imaging modalities such as CTA and TOF-MRA are comparable to DSA in detecting luminal stenosis. ²⁹ HR-VWI not only detects more plaques than DSA, ²⁹ but also achieves better inter-rater agreement. Mazighi et al. analyzed the prevalence of ICAD in patients with fatal stroke. ⁴ Intracranial plaques occurred in 62% of patients, and only 63% of these plaques (luminal stenosis >30%) were identified on conventional luminal imaging. ⁴ Both TOF-MRA and CTA are ubiquitously used as non-invasive and readily available imaging modalities in detecting culprit plaques. However, when compared to HR-VWI, these modalities provide limited information about the vessel wall, plaque contrast enhancement, plaque morphology, PB, and RI. These plaque characteristics are important determinants of plaque's stability.^18,30 Based on this observation and others, 3T HR-VWI should be the default imaging modality for diagnosis and characterization of symptomatic ICAD patients. As a non-invasive imaging technique that does not use ionizing radiation, HR-VWI may offer improved plaque characterization than conventional luminal imaging techniques.

The main limitation of this study is that not all patients underwent every imaging modality. Also, the time from stroke event to imaging varied between patients and modalities. Therefore, we focused our analysis on plaque characteristics such as degree of stenosis, PB and RI, rather than variables known to change over time, such as intraplaque-hemorrhage and contrast enhancement.^31,32 In addition, since part of the inclusion criteria was stroke/TIA due to a culprit plaque, the independent raters could have been biased towards detecting plaques. However, the criteria for interrater agreement on correct plaque detection was based on the accurate localization of the culprit plaque and not just its detection. Medical treatment such as statins may also affect plaque remodeling due to different interval time between the several imaging modalities. We also acknowledge that 7T HR-VWI is not readily available for clinical use. However, the main aim of this study is to compare different standard imaging modalities with HR-VWI.

Conclusion

The degree of stenosis as currently measured on conventional imaging based on WASID criteria does not reflect the true extent of atherosclerosis. The determination of PB and RI through HR-VWI using ROI sampling on orthogonal views, is a more accurate tool for intracranial plaque detection and characterization, and subsequent assessment of disease burden. HR-VWI should be the default imaging modality for diagnosis and characterization of atherosclerotic plaques in symptomatic patients.

Image and data acquisition was performed using an MR imaging instrument funded by 1S10RR028821-01.

Nonstandard abbreviations and acronyms

7T HR-VWI

7 Tesla high resolution vessel wall imaging

CTA

Computed tomography angiogram

TOF MRA

Time-of-flight magnetic resonance angiography

PB

Plaque burden

RI

Remodeling index

TOAST

Trial of ORG 10172 in acute stroke treatment

ASCOD phenotyping

(A, atherosclerosis; S, small vessel disease; C, cardiac pathology; O, other causes; and D, dissection)

ROI

Region of interest

WASID

Warfarin-Aspirin Symptomatic Intracranial Disease trial