Computed Tomography Texture Analysis of Carotid Plaque as Predictor of Unfavorable Outcome after Carotid Artery Stenting: A Preliminary Study

Abstract

Novel biomarkers are advocated to manage carotid plaques. Therefore, we aimed to test the association between textural features of carotid plaque at computed tomography angiography (CTA) and unfavorable outcome after carotid artery stenting (CAS). Between January 2010 and January 2021, were selected 172 patients (median age, 77 years; 112/172, 65% men) who underwent CAS with CTA of the supra-aortic vessels performed within prior 6 months. Standard descriptors of the density histogram were derived by open-source software automated analysis obtained by CTA plaque segmentation. Multiple logistic regression analysis, receiver operating characteristic (ROC) curve analysis and the area under the ROC (AUC) were used to identify potential prognostic variables and to assess the model performance for predicting unfavorable outcome (periprocedural death or myocardial infarction and any ipsilateral acute neurological event). Unfavorable outcome occurred in 17/172 (10%) patients (median age, 79 years; 12/17, 70% men). Kurtosis was an independent predictor of unfavorable outcome (odds ratio, 0.79; confidence interval, 0.65-0.97; p = 0.029). The predictive model for unfavorable outcome including CTA textural features outperformed the model without textural features (AUC 0.789 vs. 0.695, p = 0.004). In patients with stenotic carotid plaque, kurtosis derived by CTA density histogram analysis is an independent predictor of unfavorable outcome after CAS.

Keywords: brain; carotid artery disorders; computer software applications; stents; vascular accident.

Figure 1

The diagram shows the patient enrollment flowchart. Abbreviations: CAS, carotid artery stenting; CTA, computed tomography angiography.

Figure 2

Symptomatic mixed plaque of the right carotid bifurcation extended to the proximal internal carotid artery: (a) the axial computed tomography angiography image shows the internal carotid artery plaque (arrow) that determined 70% lumen stenosis, following North American Symptomatic Carotid Endarterectomy Trial (NASCET) criteria [28]; the whole plaque was segmented manually with the open-source 3D Slicer software (https://www.slicer.org/, accessed on 1 October 2021) obtaining volumetric ROI as showed in (b) axial, (c) sagittal, and (d) coronal multiplanar reformatted reconstruction derived from computed tomography angiography (arrows).

Figure 3

Diagnostic performance for prediction of unfavorable outcome (any ipsilateral neurologic acute event and periprocedural myocardial infarction or death) after carotid artery stenting. The model including textural features derived by density histogram of the plaque at CTA (blue line) outperformed the model without textural features (green line) with an area under the receiver operating characteristics curve of 0.789 vs. 0.695 (p = 0.004).