Computed tomography analysis of vulnerable carotid atherosclerotic plaque and relationship to clinical characteristics

Introduction

The World Health Organization (WHO) identified cardiovascular diseases as the leading cause of death globally [1]. Cardiovascular diseases account for most noncommunicable diseases (NCDs) deaths, or 17.9 million people annually, followed by cancers (9.3 million), respiratory diseases (4.1 million), and diabetes (1.5 million) [1]. Stroke alone accounts for an estimated 6.2 million and was ranked as the second leading cause of global death [1]. Large, randomised trials including the North American Symptomatic Carotid Endarterectomy Trial (NASCET) and the European Carotid Surgery Trial (ECST) demonstrated that carotid revascularisation reduces stroke risk in patients with symptomatic severe carotid stenosis (> 70%) [2,3], but has marginal impact in several sub-groups of patients with moderate and mild stenosis [4].

Today, carotid luminal stenosis - measured with digital subtraction angiography (DSA), computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound (US) - represents the primary determinant of treatment for patients with carotid artery stenosis. However, histopathologic studies showed that carotid plaques with identical luminal stenosis were found to be histologically different suggesting that plaque composition may trigger the increased risk of ischaemic event [5,6]. For this reason, vessel wall imaging has been increasingly used to assess plaque beyond luminal stenosis to identify those patients who may be at higher risk of cerebrovascular ischemia [7,8,9].

In the settings of ischaemic stroke, CT Angiography (CTA) is indicated for carotid imaging due to its wide availability, relative low-cost and excellent inter-operator agreement [10]. In addition to accurately assessing carotid plaque wall and luminal stenosis, CTA has been recently used to investigate selected plaque biomarkers that may be associated with plaque vulnerability and that may support patient management decisions [11,12]. Using a dataset from the prospective Biomarkers Imaging Vulnerable Atherosclerosis in Symptomatic Carotid disease (BIOVASC) study [13], this study investigated the feasibility of CTA to segment and quantify selected carotid plaque biomarkers and explored their relationship with clinical characteristics and outcomes.

Materials and methods

Image analysis

CTA images were analysed by a single reader with 3 years’ experience in carotid imaging. Curved-planar reformation (CPR) images were used to navigate along the carotid artery, and the degree of each symptomatic carotid stenosis was confirmed on the correspondent CTA axial section showing the maximal luminal narrowing (Fig. 1). As per NASCET criteria, measurements of distal ICA diameters were taken beyond the bulb where the wall is parallel [2,14]. A window width and level of 800 / 300 HU respectively were used to differentiate the contrast filled lumen from the carotid vessel wall [14].

• Download figure
• Open in new tab

Figure 1.

Semi-automated analysis of left internal carotid artery (ICA) stenosis using CT angiography. A and B: Curved planar reformatted images (CPR) taken at different angles of the left ICA showing the length of the stenosis - proximal (blue) and distal (red) stenosis boundaries. C: left ICA 3D reconstruction. D and E: axial sections of the proximal (blue) and distal (red) boundaries of the ICA stenosis. F: CTA axial slice showing the point of maximum ICA stenosis.

Carotid atherosclerotic plaque was defined as the area between inner and outer boundaries of a thickened vessel. The software package “Plaque analysis” available with AquariusNet iNtuition™ (TeraRecon) was used for the semi-automatic HU-based analysis of the plaque components. Distal and proximal boundaries were drawn manually along the vessel to enable the software to perform a semi-automatic segmentation of the vessel wall from the lumen along the selected ICA volume. Manual adjustments were made on the axial plane when necessary. Consequently, three histologically validated HU ranges [15,16] were applied to obtain volumetric measurements of plaque components within the vessel wall as follows: < 61 HU for lipid, 61 to 130 HU for fibrous tissue, HU > 131 for calcium (Fig. 2).

• Download figure
• Open in new tab

Figure 2.

Curved planar reformatted CTA showing a severe left ICA stenosis. A and B: pre-defined HU range colour map applied to the ICA stenosis. C: ICA 3D reconstruction. D: luminal boundaries. E: plaque wall boundaries with HU range colour map. F: CTA axial slice showing the point of maximum ICA stenosis. Plaque view was able to identify and quantify the different plaque components: LRNC (yellow), fibrous tissue (red) and calcium (green).

Results

One-hundred and two symptomatic carotids were analysed. Atherosclerotic lesions were categorised according to NASCET criteria into four groups: mild (< 50%), moderate (50% – 69%), severe (70% - 95%) and near-occlusion (> 95%). Across the four stenosis categories, the mean volume for lipid was 170 mm³, for fibrous tissue was 140 mm³ and for calcification was 170 mm³ (Table 2).

The Mann-Whitney U test showed greater LRNC volume in moderate stenosis compared to severe stenosis (190 mm³-150 mm³, p=0.012). Kruskal-Wallis test did not show any significant variation between volumetric measurements of plaque fibrous tissue (p=0.44) and calcifications (p=0.25) across the different categories of stenosis. After adjusting for confounders such as smoking, diabetes, therapeutic treatments and degree of stenosis, multiple regression analyses were calculated to predict plaque volume based on age and gender. Results of the regression analysis showed that these variables may predict increased plaque volume F = 7.74, p<0.001 and R² = 0.37, and they both added statistically significantly to the prediction (p<0.05). Plaque mean volume was larger in males compared to females (550 mm³ vs 300 mm³, p< 0.001); moreover, a significantly larger volume of plaque LRNC was found among male patients when compared to the female group (210 mm³ vs 100 mm³, p < 0.001). A trend for reduced mean plaque LRNC was identified in patients receiving statins (130-210 mm³, p=0.08) (Table 3). The intra-reader reliability, measured with the ICC test, showed overall good overall agreement for all the HU ranges and plaque features analysed as shown in Table 4.

Discussion

In agreement with current literature, this study extends the knowledge supporting CTA as a reliable diagnostic tool capable of quantifying selected plaque components that have histologically been associated with plaque vulnerability [17]. In-vivo evaluation of carotid plaque beyond luminal stenosis is becoming a well-known technique able to provide a comprehensive assessment of plaque morphology and may become helpful in patient risk-stratification.

Carotid plaque volume in this study was significantly larger in males than females; moreover, carotid plaque in the male group seemed to be more complex compared to the female group due to their larger volume of LRNC. Although further research is needed, these findings support the evidence that in women hormones may have a key protective role in plaque stabilisation [18]. This study also identified age as an independent predictor of plaque volume increase, which correlates to the faster carotid plaque progression rate occurring in the older population [19].

In this analysis we found larger volumes of LRNC in the cohort of symptomatic patients presenting with moderate stenosis and a trend for reduced mean plaque LRNC volume was identified in patients receiving statins. The most immediate implication of such findings is that despite causing a moderate vessel occlusion, these plaques should be considered at a higher risk of causing ischaemic events such as stroke or TIA. Nevertheless, plaque biomarkers are not commonly used in clinical settings where the degree of carotid artery stenosis is still the most common criteria used to predict the risk of cerebrovascular events in patients affected by carotid atherosclerosis.

In this study we did not cross-validate the CTA findings with histological specimens, and we adopted pre-defined plaque-HU ranges that previously showed good association with corresponding plaque histology [6,16]. CTA has always been considered an accurate diagnostic test for the identification of several carotid morphological features such as luminal narrowing, intraplaque calcifications and ulcerations [6]; however, CTA plaque components characterisation mostly remains a research tool with limited application in clinical practice [20,21]. The HU values that we used for plaque soft component characterisation (<61 HU for lipid) did not discriminate attenuation differences between intra-plaque haemorrhage (IPH) and LRNC; therefore, this study limited the CTA carotid plaque analysis to the identification of only three plaque components: LRNC, fibrous tissue and calcium. The detection with CTA of other plaque components associated with plaque vulnerability such as IPH and fibrous cap (FC) is challenging and conflicting results were reported [6,22]. These plaque components have been associated with increased risk of cerebrovascular ischaemic event [23]; however, due to spatial resolution limitations and similar attenuation coefficients CTA has not been validated yet as a LRNC – IPH characterisation tool in clinical practice. The current literature acknowledges MRI as the imaging modality of choice in the detection of LRNC [6,7]; however, low availability, high cost and long scanning time are some of the limitations of carotid MRI in acute stroke settings. For these reasons we believe that further studies on large cohorts of patients should be supported to confirm the potential of CTA as a plaque characterisation tool.

The recent introduction of Photon-Counting Computed Tomography (PC-CT) scanners in clinical practice may have the potential to significantly change clinical CTA and plaque imaging. Briefly, PC-CT relies on detectors that do not require a separate step to convert the X-rays into light [24]. Each photon that hits the detector of a PC-CT system produces an electrical pulse with a height proportional to the energy deposited by the photon; therefore, PC-CT can count the number of incoming photons and measure their individual energy [24]. This technique results in improved spatial resolution, higher contrast-to-noise ratio, correct beam-hardening artifacts, lower radiation exposure and optimized spectral imaging [25]. For these reasons PC-CT may have the potential to overcome most of the current technical demands of plaque imaging associated with high-resolution requirements and image noise. We believe that further research in the settings of non-invasive high-risk plaque imaging should be performed to explore whether PC-CT may have the potential to characterize plaque IPH and LRNC separately, and to identify early sign of plaque micro-calcification which has been recently associated with increased plaque vulnerability [26].

In this study, CTA showed good reproducibility between measurements taken at different time-points. The main causes of variability between repeated measurements were represented by the manual adjustments of plaque contour and vessel lumen, and by the definition between a slightly thickened and a normal vessel. In our analysis the semi-automatic outlining of the outer and inner border of the vessel was reviewed manually to avoid overestimation of plaque LRNC volume that may have been caused by the involuntary inclusion of peri-arterial fat within the volume of interest.

CTA does not have the same limitations as US in the presence of calcium, which has been seen as a plaque feature capable of conferring more stability to the carotid plaque independently from the degree of luminal stenosis [6]. In our analysis we found that a more accurate plaque semi-automatic segmentation is achievable when non-calcified plaques are analysed. Manual adjustment necessary to correct the initial semi-automatic plaque outlining were necessary in 24% of the cases. In these cases, the plaque algorithm applied to the imaging dataset automatically identified both the contrast-enhanced lumen and adjacent plaque calcifications (> 130 HU) as a single volume. To ensure consistency in the CTA plaque analysis and to avoid over/under estimations of calcified plaque volume, we recommend to manually confirm stenosis boundaries on the axial CTA dataset using a window width / window level of 800 / 300 HU, which allows to optimally differentiate the contrast-opacified lumen from plaque calcifications [15]. Therefore, we suggest the use of a dedicated CTA carotid plaque protocol in which the acquisition settings are accurate for both plaque depiction and patient radiation dose safety.

Finally, this study demonstrated that in-vivo MDCT assessment of plaque volume and components is feasible with a good intra-reader reliability. Moreover, the results of this study suggest that increased plaque volume may be independently predicted based on the age and gender of the patient. This study showed that larger accumulations of LRNC were found in the moderate stenosis group compared to other NASCET plaque categories suggesting that CTA-HU measurements can facilitate the immediate quantification of key components of carotid atherosclerotic plaque; whose presence, or absence, is linked to plaque vulnerability.