Towards autonomous collision avoidance in surgical robots using image segmentation and genetic algorithms

Surgical robots provide various advantages, like improved precision and dexterity, over conventional surgical procedures and have been extensively studied in the last decade. However, safety is still a major concern mainly due to the unreliable or no haptic feedback to the surgeon and the shaky movements (tremors) in the surgeons hands. These issues may result in unwanted collisions between the robotic arm tips and the soft tissues of the human body or between the robotic arm tips inside the human body causing non-repairable damage. In this paper, we propose to overcome this problem by using image segmentation techniques, on the real-time video of the robotic surgery, along with genetic algorithms to avoid possible collisions. The main idea is to introduce an automatic collision avoidance mechanism upon detecting a threat for a possible collision. For illustration, some preliminary results on using image segmentation on a video of the cholecystectomy surgical procedure are presented.