Improving cataract surgery procedure using machine learning and thick data analysis

Abstract

Cataract surgery is one of the most frequent and safe Surgical operations are done globally, with approximately 16 million surgeries conducted each year. The entire operation is carried out under microscopical supervision. Even though ophthalmic surgeries are similar in some ways to endoscopic surgeries, the way they are set up is very different. Endoscopic surgery operations were shown on a big screen so that a trainee surgeon could see them. Cataract surgery, on the other hand, was done under a microscope so that only the operating surgeon and one more trainee could see them through additional oculars. Since surgery video is recorded for future reference, the trainee surgeon watches the full video again for learning purposes. My proposed framework could be helpful for trainee surgeons to better understand the cataract surgery workflow. The framework is made up of three assistive parts: figuring out how serious cataract surgery is; if surgery is needed, what phases are needed to be done to perform surgery; and what are the problems that could happen during the surgery. In this framework, three training models has been used with different datasets to answer all these questions. The training models include models that help to learn technical skills as well as thick data heuristics to provide non-technical training skills. For video analysis, big data and deep learning are used in many studies of cataract surgery. Deep learning requires lots of data to train a model, while thick data requires a small amount of data to find a result. We have used thick data and expert heuristics to develop our proposed framework.Thick data analysis reduced the use of lots of data and also allowed us to understand the qualitative nature of data in order to shape a proposed cataract surgery workflow framework.