Automatic purse-string suture skill assessment in transanal total mesorectal excision using deep learning-based video...

Area of Science:

• Surgical skill assessment
• Medical artificial intelligence
• Endoscopic surgery

Background:

• The purse-string suture is critical in transanal total mesorectal excision.
• Objective assessment of surgical skills is essential for training and quality improvement.

Purpose of the Study:

• To develop an automated skill assessment system for purse-string suture using deep learning.
• To evaluate the reliability and correlation of the AI-generated scores with manual assessments and surgical experience.

Main Methods:

• Deep learning model trained on manually scored videos of purse-string suturing from transanal total mesorectal excision.
• AI-based image regression analysis to predict skill scores.
• Spearman's rank correlation coefficient used to assess correlations between AI scores, manual scores, suture time, and surgeon experience.

Main Results:

• The AI system demonstrated feasibility, with a low mean absolute error between AI and manual scores (0.42).
• AI scores showed significant correlation with purse-string suture time (r = -0.728) and surgeon experience (P<0.001).
• The developed AI system for skill assessment proved reliable.

Conclusions:

• An automated deep learning-based system for assessing purse-string suture skills in transanal total mesorectal excision is feasible and reliable.
• The AI score serves as a dependable metric for evaluating surgical proficiency.
• Potential for expansion to other endoscopic procedures and surgical skill assessment applications.