An Ensemble Learning Approach for Automatic Brain Hemorrhage Detection from MRIs

Brain hemorrhage is one of the conditions that could affect people for several reasons such as high blood pressure, drug abuse, aneurysm, and trauma. Neurologists ordinarily use Magnetic Resonance Imaging (MRI) scan to examine patients for brain hemorrhage. In this study, we have developed an intelligent automatic model to identify MRIs of patients with hemorrhage from intact ones using an ensemble learning approach. Moreover, our proposed model can annotate the affected area of the brain in an axial view of MRI, which helps trainee doctors to improve their reasoning and decision making. In our experimental settings, we have applied a segmentation-based feature texture analysis to prepare MRIs for classification using an adaptive boosting algorithm. Our proposed method has achieved a classification accuracy of 89.2%, with 100% sensitivity in detecting the affected area of the brain.