Artificial intelligence classifies surgical technical skills in simulated laparoscopy: a pilot study

Area of Science:

• Surgical Education
• Artificial Intelligence
• Computer Vision

Background:

• Objective assessment of surgical skills is crucial for competency-based medical education.
• Current advancements in computer vision have not been fully integrated into surgical training.
• Laparoscopic surgery requires specialized technical skills that necessitate objective evaluation methods.

Purpose of the Study:

• To develop and evaluate a computer algorithm for the automatic classification of surgical skills in laparoscopy.
• To assess the feasibility of using artificial intelligence (AI) for objective evaluation of surgical expertise.
• To explore the potential of AI in enhancing surgical training and certification processes.

Main Methods:

• A prospective study utilized a bench-top simulator for the "precision cutting" task from the Fundamentals of Laparoscopic Surgery.
• An AI algorithm, based on a transformer neural network, was developed to classify laparoscopic task videos as expert or novice.
• The model underwent fivefold cross-validation, with performance evaluated using standard metrics including accuracy, sensitivity, specificity, and AUC.

Main Results:

• The AI model achieved an accuracy of 0.867 in classifying novice versus expert surgical skills based on video analysis.
• Classification accuracy was independent of task completion time.
• A lower accuracy of 0.57 was observed for single-image classification.

Conclusions:

• This pilot study demonstrates the potential of AI and computer vision for objective classification of surgical expertise.
• The developed algorithm shows promise for supporting and enhancing surgical training and certification.
• Further validation is required to integrate this technology into routine surgical education and assessment.