Enhancing anatomical recognition by surgeons during pelvic lymph node dissection using artificial intelligence

Area of Science:

• Surgical anatomy
• Medical artificial intelligence
• Surgical education

Background:

• The lateral pelvis is complex and variable, posing challenges for pelvic lymph node dissection (PLND).
• Accurate identification of anatomical structures is crucial for safe and effective PLND in colorectal, gynecological, and urological surgeries.

Purpose of the Study:

• To develop an artificial intelligence (AI) model for identifying key anatomical structures relevant to PLND.
• To evaluate if AI assistance improves surgeons' recognition of pelvic anatomy during PLND.

Main Methods:

• An AI model was trained on over 23,000 images from 293 PLND videos.
• 36 surgeons reviewed 640 video snippets, with and without AI assistance.
• Performance was evaluated using Dice similarity coefficients and surgeon sensitivity/specificity.

Main Results:

• The AI model achieved high Dice similarity coefficients for key structures (e.g., obturator nerve: 0.8654, external iliac vein: 0.8736).
• AI assistance significantly improved surgeon sensitivity and specificity in identifying pelvic anatomical features (p < .001).
• Improvements were observed across different surgical specialties and surgeon experience levels.

Conclusions:

• The developed AI model shows potential for assisting surgeons in identifying critical pelvic anatomy during PLND.
• AI-assisted surgical navigation may enhance precision and safety in complex pelvic surgeries.
• Further research in continuous intraoperative settings is needed to confirm clinical impact.