Early Detection of Stroke using Texture Analysis

Abstract

Ischemic brain stroke has been normally examined through Computed tomography (CT) images to decide about the further treatment of the patient. However, in the acute phase, detection of the lesions may be difficult through regular visual analysis of CT images. This paper presents a method to distinguish normal tissue and affected areas by texture analysis of CT images. Five regions of interest (ROI) are selected from area that may be potentially affected by ischemic stroke, and Five from the unaffected area per image are selected for calculation of 22 texture parameters. All ROIs for each subject were classified by an expert radiologist. The ratio of all the texture features from both areas is analyzed, and finally this study suggests that texture analysis could be a useful tool to help neurologists in the early detection of ischemic stroke. The novelty of the method is that the algorithm is based on the ratio of only five texture features to classify the CT image. The preliminary results show that the accuracy of the algorithm is 93.3%.