Computer Vision News - March 2021

Projects by RSIP Vision 10 Current challenges in ICH healthcare ICH occurs when a blood vessel erupts within the cranium. Hemorrhage can be detected in CT or MRI scans using contrast injection. The bleeding is often followed by edema, which is more difficult todetect as it appears as a subtle dark area around the hemorrhage. This causes a “mass-effect”, a state where areas of the brain are compressed due to the limited volume inside the skull. Increased pressure can cause damage to brain tissue and structures, resulting in brain injury, intubation, mechanical ventilation, or even death. To avoid this undesired outcome, it is essential to diagnose ICH quickly and accurately . Standard care for ICH requires a brain scan, followed by a series of scan readings by technician, radiologist, and intensive care physician. This process can take a long time, which can lead to patient complications. Reducing time from scan to diagnosis can improve outcome and save lives. Automatic ICH detection and assessment RSIP Vision has developed an alternative solution. Automatic Detection and Assessment of Intracranial Hemorrhage (ICH) segmentation of the bleeding area and the adjoined edematous zone is feasible using advanced AI and computer vision techniques. Additional segmentation of the brain ventricles can indicate structural damage, and even provide a map for procedural planning. However, brain segmentation is not elementary. Advanced AI for ICH and edema segmentation Both bleed and edema may have a variety of shapes, and image contrast may not be sufficient. This segmentation task requires a large, annotated dataset for training, which is not easy to obtain. We therefore developed a hybrid approach where we used classic computer vision algorithms including superpixel and "… fast, robust, and accurate segmentation of multiple ICH type."