IC-RCM: Intraoperative Real-Time Detection of 5D RCM-Misalignment in Robot-Assisted Ophthalmic Surgeries with a Single-Camera System

Abstract

Robot-assisted ophthalmic surgery (RAOS) advancements have transformed precision in eye surgical procedures. RAOS offers unparalleled accuracy and dexterity, allowing for intricate maneuvers in delicate ocular structures with minimized invasiveness. However, ensuring the reliability and quality of the instruments' pivot point at the trocar, called Remote Center Motion (RCM), is imperative to uphold patient safety and surgical efficacy. Hence, real-time monitoring and tracing of RCM performance parameters are pivotal in minimizing surgical trauma and averting adverse outcomes. Our approach showcases a precise pipeline for recognizing RCM misalignment by integrating the robot's kinematic software data with neural network-based pose estimation methodologies with the mean absolute rotation error of 5.48, tested on a 23G surgical cannula with a diameter of 0.640 mm. This method uses a high-resolution camera at 4024 × 3036 capable of processing at 20 frames per second.