Real-time video enhancement for the removal of surgical lighting artifacts in computer-assisted orthopedic surgery

Abstract

Optical camera systems are commonly used to monitor patients in computer-assisted orthopedic surgery, but strong and flickering surgical lights interfere with this imaging and degrade video quality. While algorithms exist to remove the surgical light's flicker and spotlight artifacts, no algorithm exists that can handle both simultaneously. We therefore propose the first such algorithm for the joint removal of these artifacts. The algorithm works by iterating between the two artifacts, modelling their effects and correcting for them as we go. The resulting algorithm was validated on video from a mock percutaneous scaphoid fixation surgery by measuring the success of a downstream hand pose detection task. Results showed that our video enhancement, running in real-time, led to a 82-314% increase in successful hand pose detections compared to correcting for only one of the artifacts present in the image.