Artificial Intelligence-Based Surgical Phase Analysis Enables Objective Assessment of Surgeon Skill in Robotic Distal...

Area of Science:

• Robotic surgery
• Artificial intelligence in medicine
• Surgical workflow analysis

Background:

• Automating surgical video analysis with AI offers objective workflow characterization.
• Manual assessment of surgical procedures is time-consuming.
• AI can enhance surgical training and performance evaluation.

Purpose of the Study:

• To develop an AI model for surgical phase recognition in robotic distal gastrectomy (RDG).
• To construct a multicenter dataset of annotated RDG videos.
• To explore AI-driven metrics for characterizing surgeon experience-based performance differences.

Main Methods:

• An image classification model was trained on 137 RDG videos from 15 institutions.
• Experienced surgeons annotated nine surgical phases frame-by-frame.
• Model performance was evaluated using accuracy, precision, recall, and F1-score; phase durations were compared by surgeon experience.

Main Results:

• The AI model achieved 87.0% accuracy in recognizing nine surgical phases.
• AI analysis showed shorter durations for right-sided greater curvature lymphadenectomy in high-volume surgeons.
• Phase-specific operative performance differences were identified based on surgeon robotic case volume.

Conclusions:

• AI demonstrates robust performance for multicenter surgical phase recognition in RDG.
• AI-based analysis objectively profiles operative performance related to surgeon experience.
• This approach supports performance profiling and educational tools in robotic surgery, though generalizability requires further validation.