Computer Vision News - October 2020

RSIP Vision's Medical Projects 12 Supporting surgical robotics by combining Artificial Intelligence (AI), computer vision, and classical image processing in a single solution. Yehiel Shilo is VP R&D at RSIP Vision. He earned his PhD from the Hebrew University of Jerusalem. In recent years, we at RSIP Vision have been involved in a diverse set of projects in the field of surgical robotics . In this review we would like to share with you an overview of the field, as well as some specific examples of AI, computer vision, and classical image processing use cases. Introduction AI and computer vision play a central role in surgical robotics. Adoption and integration of these technologies remains a key challenge for the industry. In current surgeries, the eye-hand coordination of the surgeon is a key- element in the success of the procedure. When developing a surgical robotics system, this coordination is enhanced by imaging-robotic coordination. In order to enhance the human eye, multiple imaging systems are used. Each procedure has its unique requirements and hardware. In some cases, it can be a monocular or stereo gray/RGB camera; in other cases, depth camera or fluoroscopic images are used. AI algorithms have proven to be the most powerful framework for enhancing the human eye and are used in various tasks such as segmentation, classification, and feature tracking. Surgeon eye-hand coordination is enhanced by algorithms that can calibrate the hardware, register between the different coordinate systems, and track the movement of the robot in the space. These algorithms are based both on classical computer vision as well as AI methods. On top of this, a temporal understanding is required to sense the current state of the procedure. Robotic surgical systems face unique challenges. The precision in these procedures is crucial, and to achieve the precision requirements, a detailed as well as high level view is needed to check both the entire design and the integration of all the components of the system. An additional challenge in these AI for Surgi cal Robot i cs Yehiel Shilo