Photo Credit: FotografieLink
The following is a summary of “Intraoperative Applications of Artificial Intelligence in Robotic Surgery: A Scoping Review of Current Development Stages and Levels of Autonomy,” published in the December 2023 issue of Surgery by Vasey, et al.
For a study, researchers sought to find surgical artificial intelligence (AI) uses for robotic surgery that is still being worked on and to group them into four groups based on their purpose, level of autonomy, stage of growth, and type of result that was measured. AI-based apps could change robotic surgery, which has been built on a master-slave model up until now. But there needed to be a summary of this technology’s current amount of liberty or stage of progress that can be found. The search terms were put into MEDLINE and EMBASE from January 1, 2010, to May 21, 2022. Two reviewers independently worked on the abstract screening, full-text review, and data extraction.
The Yang and coworkers’ classification and stage of growth, along with the idea, development, evaluation, assessment, and long-term follow-up framework, were used to set the amount of liberty. 129 studies were used for the review. 77 studies (75%) talked about applications that gave Robot Assistance (level 1 autonomy), 30 studies (23%) talked about applications that gave Task Autonomy (level 2 autonomy), and 2 studies (2%) talked about applications that gave Conditional autonomy (level 3 autonomy).
There were no clinical studies on people because all of the studies were at stage 0 (Idea, Development, Evaluation, Assessment, and Long-term follow-up). One hundred sixteen (90%) did experiments in silico or outside of living things on inorganic materials, nine (7%) did experiments outside of living things on organic materials, and four (3%) did experiments inside living things using models. The study of AI applications used during robotic surgery is still in its early stages, and most of these applications have limited freedom. As systems become more autonomous, the review focus changes from AI-specific measures to process results. However, systems need to be able to be compared using shared standards.
