Impact of Artificial Intelligence on the Timing of Recurrent Laryngeal Nerve Recognition during Robot-Assisted Minimally Invasive Esophagectomy.

Background: As a first step to prevent recurrent laryngeal nerve (RLN) palsy, we have developed an artificial intelligence (AI)-based anatomical recognition system for critical anatomical structures in robot-assisted minimally invasive esophagectomy (RAMIE). In the present study, we investigated whether AI would enable surgeons to rapidly recognize the RLN.

Patients And Methods: Five surgical videos of RAMIE were used to validate the AI. The confidence level (CL) was established as a new criterion to define how confidently surgeons recognize the RLN and stratified into three levels: CL0 is the level when looking for RLN candidates; CL1, after recognizing a candidate but before confirming it as the RLN; CL2, after confirming that the candidate is the RLN. Eight trainee surgeons watched the original and AI-enhanced videos with an interval of > 4 weeks, and they were instructed to declare the RLN's location at the start of CL1 and CL2. The time to CL1 and CL2 from the beginning of RLN lymph node dissection with and without AI were compared.

Results: In all cases, the average time to CL1 and CL2 of the right and left RLN recognition was reduced by using AI. Particularly, in the right RLN recognition, significant differences were found between the surgeons using and not using AI (CL1, 134 vs. 178 s, p < 0.001; CL2, 233 vs. 325 s, p < 0.001).

Conclusions: This study demonstrated that AI would enable surgeons not only to rapidly identify the RLN but also to enhance their confidence in its identification.