LAUSR

Ischemic brain stroke remains a global health concern and a leading cause of mortality and long‐term disability worldwide. Despite significant advancements in acute stroke management, the incidence and burden of this devastating cerebrovascular event continue to increase, particularly in developing nations. This study proposes a novel machine learning approach for classifying brain stroke Computed Tomography (CT) images into its subtypes using an efficient feature descriptor. The presented descriptor is a Modified Weber Local Descriptor (MWLD), which incorporates the structure tensor for precise orientation computation and a multi‐scale approach to capture multi‐resolution features. Further, analysis of variance ranking for discriminative feature selection was applied to the MWLD features. These ranked features were tested on 4850 CT images (i.e., 875 acute, 1447 chronic, and 2528 normal) using various classifiers, such as the nearest neighbor classifier and ensemble models. The methodology achieved 98.34% (highest) testing accuracy with a fine k‐nearest neighbor classifier, outperforming existing descriptors. The MWLD descriptor and machine learning technique can accurately diagnose ischemic stroke, enabling improved clinical decision support.