Surgical video-based temporal action analysis algorithm and competency assessment in laparoscopic cholecystectomy:...

Area of Science:

• Surgical Education and Training
• Medical Image Analysis
• Computational Pathology

Background:

• Laparoscopic cholecystectomy (LC) is a core surgical procedure, but current competency assessments are subjective and resource-intensive.
• Existing methods often focus on outcomes, neglecting process-oriented metrics like surgical action timing.
• There's a need for objective, scalable tools to evaluate surgical proficiency.

Purpose of the Study:

• To develop and validate a novel algorithm for automated, objective assessment of surgical competency in LC.
• To integrate process-oriented metrics (temporal action patterns) with outcome-oriented metrics for enhanced prediction.
• To establish a scalable and interpretable solution for surgical performance evaluation.

Main Methods:

• Utilized the Cholec80 dataset of 80 LC videos.
• Employed a lag-aligned distance method for stage-specific temporal action segment comparison.
• Developed a support vector regression (SVR) model incorporating SHapley Additive exPlanations (SHAP) identified features.

Main Results:

• Cumulative temporal dissimilarity scores significantly correlated with ground-truth competency scores (r = -0.61, P < .001).
• The SVR model, using critical-view-of-safety score, operative time, and temporal dissimilarity, achieved high predictive accuracy (r = 0.86, P < .001).
• The model effectively predicted surgical competency scores based on video analysis.

Conclusions:

• A novel process-oriented algorithm for video-based temporal action analysis was introduced.
• The integrated model provides accurate, quantitative prediction of surgical competency.
• This automated approach offers a scalable and objective advancement over traditional surgical assessment tools.