Comprehensive Summary
Zhao et al present an AI deep-learning model that can screen Robot-assisted radical prostatectomy (RARP) videos and score them based on a set of parameters in order to rank the surgeon's expertise. Between August 2016 and April 2024, 410 RARP surgery videos were collected from 18 education centers in Japan. Of those, 213 videos were selected and categorized into training, validation and testing sets. These videos were analyzed in different stages of surgery in order to form a set of key variables which allowed the AI model to rank surgeons from expert to novice. After all selected videos were annotated and a bias test was performed it showed that a high AI confidence score(ACIS) was a sign of higher expertise. The final AI scoring system had an accuracy of 86.2% in being able to differentiate between experts and novices. Currently, assessment of robot assisted prostate surgeries is done manually which is time consuming and labor intensive. Since the later stages of surgery are the most complex, analyzing the amount of time they take is a way to assess the proficiency of the surgeon. However some facilities may only have data for novices or for experts which may show bias in the resulting statistics.
Outcomes and Implications
The current way of assessing surgeries is prone to bias, as it is done by humans who must watch videos over and over again. An AI model will be able to go through and automatically be able to set parameters for itself and evaluate surgeons on that basis. It will also be much more cost and labor effective. It will also allow for standardization across institutions as the same AI could be implemented in multiple places at once, allowing for easier comparison center to center. These factors combined will allow for a much better system of scoring and feedback for surgeons, which in turn will create better patient experiences and further outcomes.