An Intelligent Approach for Automating Robotic Arm Maneuvering in Endometriosis Surgery

Artificial intelligence (AI) and computer vision are revolutionizing numerous fields, including robotic surgery, which stands to benefit immensely from advances in machine learning methodologies. While prior research has extensively focused on disorder detection, localization, and semantic segmentation, the crucial challenge of robotic arm maneuvering during autonomous surgeries remains underexplored. This study proposes a robust and interpretable approach to enable robots to autonomously execute endometriosis surgeries by skillfully navigating their arms, equipped with a camera and surgical tools such as graspers or lasers. A decision tree framework is developed to assess the robot's real-time status and guide its actions at every surgical stage. This framework integrates diverse ensemble neural network models for classification and semantic segmentation to support decision-making. Specifically, the proposed ensemble classification models utilize deep learning to assess image quality, identify obstructions caused by adhesions, detect anatomical targets (e.g., uterus or peritoneum), and determine the proximity of the ovary to the uterus. The proposed ensemble semantic segmentation models further enhance accuracy by detecting and localizing the uterus and ovary. By employing these ensemble frameworks within the proposed decision tree model, this work aims to advance robotic surgery capabilities, enabling fully autonomous, reliable, and efficient operations. Consequently, the proposed method aims to minimize economic costs, bleeding, post-operative pain, and infection risk, while optimizing surgical precision and performance.