Computational Intelligence Approach for Predicting Ischemic Stroke using Brain MRI

Stroke occurs when the vessels that supply blood to the brain either blocked by a clot or ruptures. Magnetic resonance imaging (MRI) is widely used in diagnosis process of stroke, assist physicians in finding the etiology for the incident, and to predict the functional outcome. Computer-aided diagnosis is essential for fast, accurate diagnosis and monitoring the progress of the disease. In this paper, we propose an automated process to segment the stroke lesions in diffusion-weighted imaging (DWI) modality of brain MRI. The lesion area was segmented from brain images using expectation-maximization (EM) algorithm. The statistical features were extracted and classified for stroke type according to The Oxfordshire Community Stroke Project (OCSP) using Random Forest (RF) classifier. The proposed method was implemented in 42 MR images having stroke lesions collected from hospital. The method efficiently segmented the stroke lesion and the result was evaluated with measured parameters that predicts the accuracy of the classifier. It achieved the sensitivity of 0.804, accuracy 0.835, dice coefficient of 0. 809 and Jaccard index of 0.702, which are better than to earlier published results.