Video-based formative and summative assessment of surgical tasks using deep learning

Area of Science:

• Medical technology
• Artificial intelligence in surgery
• Surgical education

Background:

• Current surgical skill assessment methods, primarily video-based assessment (VBA), are manual, subjective, and time-consuming.
• Existing VBA lacks objectivity and consistent inter-rater reliability, hindering effective surgical training and credentialing.
• The need for objective, automated, and efficient surgical skill evaluation is critical for improving clinical outcomes.

Purpose of the Study:

• To develop and validate a deep learning (DL) model for automated surgical skill assessment from video feeds.
• To enable both high-stakes summative and low-stakes formative assessments of surgical performance.
• To establish a quantitative and reproducible method for evaluating surgical tasks, enhancing surgical training and certification.

Main Methods:

• Implementation of a deep learning model utilizing video data for surgical skill analysis.
• Development of heatmap-generated formative assessments based on visual features correlating with surgical performance.
• Validation of the DL model for objective and automated skill evaluation.

Main Results:

• The proposed DL model provides automatic and objective assessment of surgical skill execution from video.
• Formative assessment is generated via heatmaps, offering insights into performance-related visual features.
• The model demonstrates potential for quantitative and reproducible surgical task evaluation.

Conclusions:

• The developed deep learning model offers a solution for objective, automated, and efficient surgical skill assessment.
• This technology can significantly enhance surgical training, certification, and credentialing processes.
• The DL model facilitates the broad dissemination of reproducible surgical skill evaluation.