PhacoTrainer: Implementing and Evaluating an Artificial Intelligence Tool to Improve Cataract Surgery
Funded Grant
Overview
Affiliation
View All
Overview
description
Project Abstract Developing cataracts is an inevitable part of aging; thus, cataract surgery is the most commonly performed surgery in the United States. Although cataract surgery can effectively restore vision, surgical training remains challenging for this delicate microsurgery. Poor instrument control or prolonged surgical time can lead to complications, sometimes resulting in permanent vision loss. Current training relies heavily on one-on-one feedback from attending surgeons, which can be subjective and variable, with limited capacity for objective review or large-scale performance tracking. Artificial intelligence (AI) has the potential to transform surgical education by automating performance assessment and feedback from video data. Our team has developed deep learning models capable of detecting 13 major surgical steps, 8 classes of surgical instruments, and key ocular anatomy in cataract surgery videos with high accuracy. From these, we generate detailed performance metrics by tracking the phases of surgery and the motion of surgical instruments and eye anatomy (including path length traversed). To deliver these insights to trainees and educators, we created PhacoTrainer (phacotrainer.com), a novel web-based platform that runs our AI models on uploaded surgical videos and presents structured, interactive, visual feedback on surgical performance. PhacoTrainer enables users to track skill development longitudinally and automatically helps identify areas for improvement, addressing a major unmet need in surgical education. Our long-term goal is to build tools that enhance ophthalmic surgical training and patient outcomes. In this project, we propose a staged implementation study of PhacoTrainer AI metrics across three ophthalmology residency training programs in the United States. We aim to (1) validate PhacoTrainer AI metrics across residency programs, establishing normative ranges and learning curves for AI- derived performance metrics by correlating them with increasing surgical skill; (2) measure the educational impact of access to PhacoTrainer AI metrics on the cataract surgical learning curve, as assessed by masked experts using a validated surgical skill scoring system, and as assessed by trainee self-report on surgical confidence. Completion of this study will result in an optimized and updated PhacoTrainer application which delivers validated AI-generated surgical skill metrics to trainees and improves their surgical skill acquisition, ready for broader deployment. In developing AI-derived objective and quantitative performance tracking for surgery, this work represents a paradigm shift in surgical education and lays the foundation for AI-driven surgical feedback nationwide. Though PhacoTrainer was developed for cataract surgery, our approach can be generalized to other ophthalmic and non-ophthalmic surgical fields. By modernizing how we train and evaluate surgeons, we aim to ultimately improve surgical outcomes and patient care on a wide scale.
