Explainable AI-Powered Framework for Video-Based Skill Assessment in Cataract Surgery

Explainability matters for clinical AI: Unlike previous automated surgical assessment tools that operated as "black boxes," this framework prioritizes explainability by breaking down skill evaluation into ten specific, interpretable motion-based metrics that surgeons can understand and validate, rather than relying on opaque AI predictions.

AI can achieve high accuracy while remaining transparent: The framework demonstrated 87% accuracy in assessing cataract surgery skills while maintaining interpretability, suggesting that explainability and performance are not mutually exclusive—a critical finding for responsible deployment of AI in high-stakes medical settings.

Objective metrics can complement expert judgment: The study found strong correlations between automatically computed metrics and expert surgeon ratings, indicating that AI-powered assessment could reduce reliance on subjective evaluation and provide consistent, standardized performance feedback across different training programs.

Large, specialized datasets are foundational: The creation of the world's largest cataract surgery video dataset (2,000 recordings) was essential to developing reliable AI assessment tools, highlighting that responsible AI development in specialized domains requires substantial investment in high-quality, domain-specific data collection.

Automated assessment could address workforce shortages: By enabling objective, scalable evaluation of surgical skills without requiring expert surgeons to manually review every training video, this framework could help accelerate surgical training and address persistent shortages in the surgical workforce.