Computer Vision News - March 2021

RSIP Vision has developed a cutting edge deep neural network to perform a very accurate brain ventricles segmentation through advanced AI techniques. RSIP Vision improves ICH healthcare AI and computer vision can improve both diagnosis and treatment of ICH. RSIP vision’s team consists of experienced clinicians and engineers, working side-by-side to achieve the best solution for this segmentation problem. We managed to obtain great results by combining state-of- the-art deep learning algorithms with classic computer vision techniques. We can implement this method into medical devices quickly and efficiently, bringing a solution to this segmentation problem and improving ICH outcome. More articles about AI for brain and neurology . 11 Detection and Assessment of ICH graph-cuts for initial segmentation, whichallowsexperts toeasilyannotate. A limited annotated dataset is used for training a small neural network , and subsequently, we implemented a closed-loop bootstrapping solution to iteratively improve the training set, and train bigger and better neural networks. As a result, RSIP Vision’s fully automated segmentation technology achieves fast, robust, and accurate segmentation of multiple ICH types, with constant running time. The resulting segmentation is a 3D mesh from which blood volume can easily be calculated, indicating severity of the ICH. This information can immediately be transmitted to the physician, and adequate treatment can be selected. Brain Ventricle Segmentation using Deep Learning Non-invasive treatment methods are preferrable. However, sometimes surgical procedures are inevitable. Appropriate pre-planning can shorten the procedure and improve results tremendously. Segmentation of the brain ventricles in addition to the above-mentioned segmentation provides a pre-op map of the brain and allows the surgeon better preparation. "Automatic segmentation of the bleeding area and the adjoined edematous zone is feasible using advanced AI and computer vision techniques." … brain segmentation is not elementary.