Intraoperative video analysis and machine learning models will change the future of surgical training

摘要

Machine learning (ML) is an application of artificial intelligence (AI) which enables automatic learning from large datasets. The advances in this computer science coupled with the increase in minimally invasive surgery (MIS) and availability of surgical videos, has resulted in ML algorithms designed to analyse intraoperative videos. This technology aims to improve surgical training and surgical assessment resulting in improved surgical safety. A literature search of MEDLINE and EMBASE was conducted. The search terms included the following, individually or in combination: ‘machine learning, ‘video analysis’, ‘computer vision’, ‘neural networks’ and ‘surgery’. Relevant articles were scanned and included in the discussion. These include research developing ML algorithms for surgical phase recognition, instrument recognition, gestures recognition, and anatomical landmark recognition. The implications for the future of surgical training are discussed. The next decade is likely to see a huge increase in MIS, particularly robotic surgery, and ML video analytics of these operations. This is likely to enhance surgical training and reduce surgical errors. However, there is a necessity for much bigger datasets for all operative procedures to allow increasing accuracy of the ML algorithms.