Intracranial Hemorrhage Detection

Abstract

ICH is diagnosed through history, physical examination, and, most commonly, non- contrast CT examination of the brain, which discloses the anatomic bleeding location. Trauma is a common cause. In the absence of trauma, spontaneous intraparenchymal hemorrhage is a common cause associated with hypertension when found in the deep locations such as the basal ganglia, pons, or caudate nucleus. [7] Automatic triage of imaging studies using computer algorithms has the potential to detect ICH ear- lier, ultimately leading to improved clinical outcomes. Such a quality improvement tool could be used to automatically manage the priority for interpretation of imaging studies with presumed ICH and help optimize radiology work ow. Machine learning and computer vision are among a suite of techniques for teaching computers to learn and detect patterns. [18] We have to identify acute intracranial hemorrhage and its subtypes. In this problem a patient can have more than one sub type of ICH so this problem belongs to a Multilabel Classi cation Problem. We have used di erent models to classify the ICH images. 1