Edge AI-Based Intraoperative Image Segmentation for Robotic-Assisted Orthopedic Surgeries

Abstract

Robotic-assisted orthopedic surgeries have revolutionized precision in joint replacement and fracture fixation. Intraoperative image segmentation remains a significant challenge due to high computational demands and the need for real-time processing. Traditional cloud-based solutions introduce latency, security concerns, and dependency on high-bandwidth internet, making them unsuitable for time-sensitive surgical procedures. Edge Artificial Intelligence (Edge AI) offers a transformative approach by enabling on-device computation, reducing latency, and improving the efficiency of intraoperative segmentation. This paper explores the integration of Edge AI for real-time intraoperative image segmentation in robotic-assisted orthopedic surgeries. We discuss the advantages of Edge AI in reducing reliance on external servers and ensuring high-speed, accurate segmentation directly at the surgical site. The study evaluates different deep learning architectures, including U-Net, DeepLabV3, and transformer-based models, optimized for edge deployment using techniques such as quantization, pruning, and knowledge distillation. A real-time processing pipeline is proposed, integrating Edge AI hardware such as NVIDIA Jetson Xavier and Google Coral TPU to process surgical images efficiently. Experimental results demonstrate that Edge AI-based segmentation achieves real-time inference with sub-100ms latency while maintaining high accuracy. The study highlights challenges such as hardware constraints, regulatory compliance, and model generalization across different patient anatomies. We discuss future research directions, including federated learning, augmented reality integration, and improved hardware acceleration. Overall, Edge AI has the potential to enhance robotic-assisted orthopedic surgeries by providing fast, accurate, and locally processed image segmentation, improving surgical precision and patient outcomes.