Intracranial hemorrhage (ICH) represents one of the most common and severe emergencies in clinical practice, with notably high rates of mortality and disability. In the field of medical image analysis, the rapid advancement of deep learning techniques has provided robust support for ICH segmentation tasks. This study systematically reviews the primary publicly available datasets and evaluation metrics for ICH segmentation and provides a comprehensive overview of deep learning-based segmentation methods. Covered approaches include convolutional neural network (CNN) models, U-Net and its improved variants, along with other emerging deep learning techniques. The main improvement strategies of these models are summarized, and critical challenges encountered during the segmentation process are analyzed. Finally, major open issues in current research are discussed, and potential future research directions and possible solutions are outlined.