For a study, researchers sought to evaluate the performance of a deep learning (DL) system based on cataract surgery (CS) films for evaluating and supervising cataract extraction utilizing phacoemulsification with intraocular lens (IOL) implantation.
DeepSurgery was trained on 186 typical CS movies to detect 12 CS steps and was verified on 50 and 21 CS videos, respectively. The DeepSurgery guide & alert feature was evaluated using a supervision test that included 50 CS films. A real-time test with 54 CSs was also utilized to evaluate DeepSurgery grading performance to an expert panel and residents.
DeepSurgery exhibited consistent performance for all 12 recognition phases, including the length between two adjacent steps, in internal validation with an ACC of 95.06% and external validations with ACCs of 88.77% and 88.34%. DeepSurgery also detected the order of surgical steps and warned physicians of improper step orders. During the review process, six major processes are automatically and concurrently measured (centesimal system). As a result, DeepSurgery step identification performance was strong in a real-time comparison test (ACC of 90.30%). Furthermore, DeepSurgery and an expert panel evaluated the surgical stages with equivalent results (kappa ranged from 0.58 to 0.77).
DeepSurgery is a potential way to improve surgical results by providing real-time supervision and an objective surgical assessment system for regular CS.
Reference: sciencedirect.com/science/article/abs/pii/S1743919122005179
