From manual parametric to artificial intelligence-based automation: A systematic review of recent advances in endoscopic surgical skills evaluation

Evaluating surgical skills is critical to surgical training. Numerous methods and scoring scales have recently emerged to assess skills. This study systematically reviews techniques used for evaluating endoscopic surgical skills. An online search was conducted, and all the surgical skills evaluation systems pertaining to endoscopic surgeries published between 2015 and 2024 were reviewed. Based on the screening, the skills evaluation systems were classified based on the type of evaluation strategy (manual or automatic), nature of techniques for automatic evaluation (parametric-based, machine learning/deep learning [ML/DL]), country of origin, surgical sub-speciality and evaluation parameters. The study revealed 46 different surgical skills evaluation systems developed for endoscopic surgeries by different institutions. 47.8% of the articles performed manual evaluation, and 52.2% performed automated evaluation. Among automated evaluation papers, 70.8% of them performed parametric-based, 16.7% ML and 12.5% of them DL. Further, 87.5% of automatic systems used 0-10 parameters, none used more than 20. For manual methods, 54.5% used 0-10 parameters, and 22.7% used over 20. Only 17.4% evaluated generalised skills, whereas 32.6% targeted general surgery-specific skills. Endoscopic skill evaluation is shifting from manual to automated approaches. However, limited public datasets constrain automation. Manual methods assess broader skill factors, whereas automatic methods focus on fewer parameters. There is a need for automatic evaluation techniques to adopt more detailed skill evaluation parameters. Furthermore, the study reveals the need for research to evaluate the skills of various surgical sub-specialities. The study emphasises the pressing need for active research in this area and serves as a foundation.