Medical safety, especially in the operating room, is an essential concept for both patients and healthcare professionals. Our team has implemented a medical safety management artificial intelligence system using Deep Learning. The system’s performance in face recognition, left-right recognition, and intraocular lens recognition on 300 sets with 50% intentional errors was 100%. At the same time, none of the three ophthalmology residents could produce a perfectly correct answer. Although the AI pointed out the errors multiple times during left-right and intraocular lens recognition in actual clinical practice, all medical personnel in charge were unaware of the errors for more than one minute. Although the AI eventually avoided the error, these examples show how humans can make mistakes even in simple tasks.
