3D Vision Technology in Precision Medical Robots

As minimally invasive surgery (MIS) has evolved towards intelligence and automation, the need for environment perception by medical robots has shifted from mere image transmission to deep 3D scene understanding. This article first reviews the evolution of 3D vision techniques in precision medical robotics, focusing on the shift from traditional geometric methods to emerging paradigms such as deep learning, Transformer architecture, and Neural Radiation Field (NeRF). The article then delves into the special visual challenges in endoscopic scenarios, and illustrates the latest breakthroughs in dynamic reconstruction and vision SLAM techniques in the context of cutting-edge literature such as EndoNeRF and 3D Gaussian Splatting. In addition, this paper also discusses the application of these techniques in real-world scenarios such as surgical instrument tracking, multimodal navigation, and haptic perception, aiming to provide a systematic reference to support future research and development of intelligent minimally invasive surgical robotic systems.