Hyperdense Middle Cerebral Artery Segmentation in Non Contrast Computed Tomography

Abstract

The hyperdense middle cerebral artery (MCA) sign refers to focal increased density of the MCA in Non-Contrast Computed Tomography (NCCT) and is the earliest sign of acute ischemic stroke. In this paper, we present the implementation of a method that allows the automatic segmentation of the hyperdense MCA sign in NCCT pathological clinical cases, as a first phase in the development of a tool that will support the early detection of cerebral infarction. A fully automated algorithm was proposed for the delimitation of volumes of interest and the segmentation of the hyperdense MCA. Volumes of interest were defined according to the anatomical location of the suprasellar cistern, and features of the hyperdense MCA were extracted according to the Hounsfield Units and entropy. The segmentation was carried out using a model of region growing and active contours (snakes). The results show an accuracy of 96% (99% per slice) and a mean correlation of automatic versus manual segmentation of 94%.