Artificial Intelligence alert systems during robotic surgery: a new potential tool to improve the safety of the intervention

Abstract

Objective: Nowadays we are facing a raising in adverse events related to robotic procedures due to lacking well-trained robotic surgeons. To cope with this issue, along with a standardized surgical training program, the advent of new technologies can come handy in helping beginner surgeons to avoid unwanted surgical events. In this study, a model based on Artificial Intelligence (AI) was developed to predict the occurrence of intraoperative bleeding and to warn the surgeon on bleeding risk. Patients and surgical procedure: A dedicated artificial Neural Network (NN) was created and trained to recognize active bleeding during robot-assisted radical prostatectomy (RARP). The created software scans the video recorded by the endoscope every 3 seconds, and returns the confidence rating, in percentage, that the bleeding event is recognized in that interval of time. Confident rates lower than 100% were used as a score to predict the possibility of the bleeding occurrence in the next frames, alerting the medical staff beforehand. All the videos recorded were reviewed by a human operator in order to evaluate the capability of the software to correctly identify the bleeding and the performances were analyzed in terms of true/false negative/positive. Results: 10 patients undergoing RARP were enrolled from October 2021 to April 2022. Our software demonstrated: (1) Correct identification of active bleeding (true positive 98%); (2) no false alert in case of absence of bleeding (false positive 3%); (3) Ability to anticipate, by gradually increasing the %, the possible occurrence of the event; (4) Beyond the 93-95% threshold, the event occurs in the next second. Conclusion: Our software, based on NN, reveales to be able to predict the intraoperative bleeding. The prediction of adverse events during robotic surgery with Artificial Intelligence will help the beginner surgeons during interventions, potentially improving patient's safety.

Keywords

• Prostate cancer
• Artificial intelligence
• RARP
• Radical prostatectomy
• Neural networks