Delineation of infarct lesions by Multi-dimensional Fuzzy C-Means of acute ischemic stroke patients

Abstract

Lesion size in diffusion weighted imaging (DWI) of magnetic resonance (MR) images is an important clinical parameter to assess the lesion area in ischemic stroke. Manual delineation of stroke lesion is time-consuming, highly user-dependent and difficult to perform in areas of indistinct borders. In this paper we present a segmentation process to detect lesion which separates non-enhancing brain lesion from healthy tissues in DWI MR images to aid in the task of tracking lesion area over time. Lesion segmentation by Fast Fuzzy C-means was performed in DWI images obtained from patients following ischemic stroke. The lesions are delineated and segmented by Multi- dimensional Fuzzy C-Means (FCM). A high visual similarity of lesions was observed in segmented images obtained by this method. The key elements are the accurate segmenting brain images from stroke patients and measuring the size of images in pixel-wise for defining areas with hypo- or hyper-intense signals. The relative area of the affected lesion is also measured with respect to normal brain image.