Abstract:
Introduction: One of the medical specializations that produces extremely huge datasets that artificial intelligence (AI) can process in-depth and thoroughly is surgery. The phrase artificial intelligence (AI) refers to a collection of computer technologies that allow algorithms to analyze, comprehend, forecast, or act independently on clinical data. A range of statistical methods and algorithms are used in AI models that aim to replicate human cognitive processes, enabling machines to take in information from and react to their surroundings. This study was conducted with the aim of a systematic review regarding the applications of AI in surgery.
Methods: A systematic review on review studies was conducted by searching the keywords (AI and surgery), in the title, abstract in Embase, Web of Science, Scopus, PubMed scientific databases on February 1, 2024. In this study, the Guidelines and Preferred Items of Systematic Review Studies (PRISMA) were followed. Review studies that investigated the application, role and evaluation of AI in surgery and whose full text was available in English were considered as inclusion criteria without time limits.
Results: Finally, 8 studies were included in this review. In our study, the reviewed surgeries included motion analysis, urology, obstetrics, gynecology, tissue retraction, cardiac, cataract, pediatric, orthopedic, plastic, and reconstructive surgery. The quality of the reviewed studies was poor. The tiny size of the datasets limits the conclusions. Less than half of the models could be understood, and the great majority were not validated. Only a small number of published AI algorithms were impartial, interpretable, and externally evaluated. The short datasets, lack of external validation, use of algorithms without training information, and lack of use of non-specialist language to explain them to users (surgeons) are only a few of the constraints on AI research that the current review finds. According to the statistics, deep learning still has difficulties in recognizing surgical expertise and complications
bias and drift, patient safety, ethics, governance, and cybersecurity are some of the main obstacles preventing its widespread adoption. Every surgical specialty is looking into the potential of AI to maximize clinical efficiency, but a large portion of the research to date is still in the preclinical stage. Collaboration will be necessary for the integration of AI into routine clinical practice in the future. To evaluate the difficulties and guarantee accuracy and safety for use in clinical practice, further robust research must be conducted. AI can be used for diagnosis and screening in order to enable prompt treatment.
(
, 
, 
)
