Evaluation of radiomic features of carotid atherosclerotic plaques from CT angiography and integration with post-endarterectomy micro-CT imaging

Stroke is the third leading cause of death in the developed nations. Noninvasive imaging, such as Computed Tomography Angiography (CTA) is an essential tool for a better risk stratification. In this project, in vivo CTA images of carotid arteries of 33 patients selected for carotid endarterectomy (CEA) (17 symptomatic and 16 asymptomatic) are analyzed using radiomic analysis (RA) and conventional quantitative image processing (CONV); ex vivo micro-CT images of CEA samples were also acquired and analyzed. The CONV analysis showed that only one out of 11 quantitative plaque metrics (Mean HU) was significantly different between symptomatic and asymptomatic patients. For the RA analysis, 11 out of 105 features were significantly different in the two groups of patients. The ex vivo micro-CT analysis of the Ca distribution showed statistically significant differences for 11 out of 40 morphometric features of calcific microparticles. The ROC analysis showed better AUC for the RA model and RA+uCT model with respect to CONV model (CONV: AUC=0.67, 95% CI 0.63-0.71; RA: AUC=0.75, 95% CI 0.71-0.79, p=0.0056; RA+uCT AUC=0.83, 95% CI 0.79-0.86, p<10-6). The results obtained suggest that radiomic analysis is better than conventional analysis in discriminate patients with and without symptoms from standard CTA images of the carotid arteries; geometry-based quantification of microcalcifications, as obtained by micro-CT, provide additional important information which is independent from CTA.