Artificial intelligence based real-time segmentation and feature tracking in urological retroperitoneal robotic...

Area of Science:

• Robotics and Artificial Intelligence in Surgery
• Medical Imaging and Computer Vision
• Surgical Navigation Systems

Background:

• Robotic-assisted surgery, especially with the da Vinci Single-Port (SP) system, offers precision in confined spaces like the retroperitoneum.
• Retroperitoneal anatomy presents significant spatial orientation challenges for surgeons, particularly those new to the field.
• This research addresses the need for improved visualization and guidance in SP retroperitoneal procedures.

Purpose of the Study:

• To develop and validate an Artificial Intelligence (AI)-based framework.
• To enable real-time segmentation and feature tracking during SP retroperitoneal robotic-assisted nephrectomies.
• To enhance spatial awareness and surgical precision in challenging anatomical locations.

Main Methods:

• Utilized a convolutional neural network (CNN) based on YOLACT++ for instance segmentation.
• Trained the AI model on annotated video frames from 15 patients undergoing SP nephrectomies.
• Performed real-time segmentation and detection of key anatomical landmarks and surgical instruments.

Main Results:

• Achieved high Area Under the Precision-Recall Curve (AUPRC) values: 0.759-0.901 for landmarks and up to 0.887 for instruments.
• Demonstrated AI's capability for real-time analysis of surgical video feeds.
• Identified limitations including a small dataset and variable segmentation precision for certain structures like the ureter.

Conclusions:

• The AI framework shows significant potential for real-time surgical guidance in SP retroperitoneal procedures.
• This technology can potentially improve novice surgeon training and streamline complex retroperitoneal surgeries.
• Future work includes dataset expansion, accuracy improvement, and integration into live surgical environments.