Brain Intracranial Hemorrhage Segmentation using Unsupervised Learning on Volume CT Images

Payal Mohadikar, Ye Duan, Steven Carr

Abstract: In this paper we propose a novel unsupervised learning approach pipeline for brain intraparenchymal and subdural hematoma segmentation using brain volume computed tomography (CT) modality, specifically for instance when there is low contrast and noisy brain image dataset. The proposed method was analyzed using a private CT brain image dataset from a collaborator. Finally, the performance of the proposed approach was evaluated by comparing constructed segmentation results with ground truth generated by manual contour segmentation.