Advancements in Cataract Surgery Using AI Video Analysis

Recent innovations in cataract surgery are revolutionizing approaches to treatment, particularly through the integration of artificial intelligence (AI) video analysis. While manual small incision cataract surgery (SICS) remains a prevalent technique in low- and middle-income countries due to its cost-effectiveness, there has historically been a lack of publicly available surgical video datasets to support research and training in this area.

Addressing this gap, an international research team led by Sankara Eye Foundation India, in collaboration with the University Hospital of Bonn (UKB) and the University of Bonn, has developed the first AI-based system for phase detection in SICS procedures. Their groundbreaking study, recently published in Scientific Reports, introduces the 'SICS-105' dataset, comprising videos of 105 surgeries performed at Sankara Eye Hospitals in India, annotated with detailed surgical phases.

The researchers employed the deep learning model MS-TCN++, created by Prof. Dr. Jürgen Gall's team at the University of Bonn. This model can identify distinct surgical phases—such as preparation and specific steps involving the lens—with over 85% accuracy. This precise phase recognition enables quantitative comparison of surgical techniques, feedback for surgeons, and the potential for automated assessment of surgical quality, thus paving the way for improved training and surgical outcomes.

Furthermore, the team is launching the SICS-155 Challenge, expanding the dataset to include 155 annotated surgeries. This international competition aims to accelerate development of AI algorithms capable of recognizing surgical phases, ultimately supporting surgeon training and enhancing the safety and effectiveness of cataract surgeries in resource-limited settings.

While high-income countries often utilize phacoemulsification techniques, highly automated by AI, the application in SICS has been limited due to the lack of data and algorithms. This innovative work signifies a major step toward democratizing AI-supported surgical analysis globally. Future efforts also involve developing AI tools for detecting surgical instruments and complications, further advancing automated surgical video analysis.

Ultimately, integrating AI into cataract surgery processes holds promise for improving surgical precision, standardizing procedures, and increasing accessibility of quality eye care worldwide.