Symmetry Based Computer Aided Segmentation of Occluded Cerebral Arteries on CT Angiography

Abstract

Thrombus volume and length are important predictors of success of recanalization in patients with acute ischemic stroke. The lack of contrast between low density thrombus and surrounding brain tissue in CT images makes manual delineation a difficult and time consuming task. We developed and evaluated an automated lumen and thrombus segmentation pipeline on CT angiography. The method generates a shape prior from the segmentation of the contralateral artery, mirror symmetry and local image intensity. The accuracy and interobserver variability were assessed using Bland Altman analysis and the calculation of Pearson’s correlation coefficient. The accuracy of the method was comparable to the interobserver variation in manual delineations. To our knowledge, this method is the first automated thrombus segmentation on base-line CT angiography and therefore opens a way to automatic thrombus characterization for treatment planning.