Magnification and localization of vessels in robotic surgical videos based on accuracy High-Order Phase-Based Video Magnification

Robotic surgery, known for its minimally invasive benefits, faster recovery, and fewer complications, faces challenges in identifying blood vessels and neurovascular bundles during surgery, risking intraoperative injuries. Existing methods encounter limitations like alignment errors and additional hardware requirements. Addressing these challenges, we propose the Accuracy High -Order Phase -Based Video Magnification (AH-PVM) method. The process involves three stages: initially determining the main frequency of tiny pulsations using dense optical flow and k -means clustering, followed by spatial-temporal decomposition through a complex steerable pyramid for fine frequency analysis. Introducing a Gaussian third -order derivative filter magnifies high -order phase motion signals, emphasizing subtle motion features. Finally, a bilateral filter applies video time -domain filtering, removing noise, enhancing video quality, and preserving pulsation details. Experiments on vascular pulsation magnification in robotic surgical videos demonstrate improved magnification with reduced distortion and artifacts. Motion details align more closely with heart beat characteristics compared to conventional methods. Objective metrics and subjective evaluations consistently show the method's performance in terms of structural similarity, noise suppression, and error control. This work explores the feasibility and robustness of applying video magnification methods to identify vessels and neurovascular bundles in different robotic surgeries. The innovation of the method lies in achieving precise capture and enhancement of pulsations, providing a reliable computer vision tool for accurate observation of vascular micro -pulsations during robotic surgery. This approach is expected to enhance the safety and efficacy of minimally invasive surgery.