AI-Powered Framework for Cataract Surgery Video Optimization

Cataract surgery videos play a vital role in medical education and assessment of surgical skills. Preserving surgical videos is essential for medical advancements, but hospitals often struggle with storage limitations and the overwhelming number of recordings. Surgical videos also need to be properly processed before they can be effectively used as input for AI applications. This paper presents an intelligent framework that automatically optimizes the storage and preprocessing of cataract surgery videos. Using deep neural networks, we accurately extracted the position of the cornea and surgical instruments during the procedure and removed idle phases from the video as time intervals where no surgery is performed. This method significantly reduces the video file size by keeping the main information of the surgery and optimizes the storage of surgical videos. Our automated pipeline streamlines video storage management and prepares surgical videos for use in AI models, enabling surgical analysis and skill assessment. Models, and codes of this research will be available by submitting a request through this website (aras.kntu.ac.ir/ai).