Development of Multi-Center Surgical Video Benchmarks for Laparoscopic Pancreaticoduodenectomy

Background: For cross-disciplinary quality improvement needs, benchmarking is a practice-proven mature strategy that can provide reliable support for quality optimization in various domains. With the rapid development of artificial intelligence (AI), researchers have gradually begun to use AI to conduct comprehensive analysis of surgical videos. This greatly helps to standardize surgical procedures and enhance the surgical skills of surgeons. However, at present, there is no benchmark video assessment for laparoscopic pancreaticoduodenectomy (LPD). The aim of this study is to apply AI-assisted evaluation of surgical videos to establish the surgical benchmarks for LPD.

Methods: We collected multi-center LPD surgery videos prospectively to establish a video database. Then used the Surgsmart AI platform to output multi-dimensional data about these surgery videos, including the duration of each core surgical step, the usage of surgical instruments, and the number of intraoperative bleeding events. Besides, the Achievable Benchmark of Care (ABC) method was adopted to calculate the benchmark average values of various data, thereby establishing quantitative reference standards for LPD procedures.

Results: A total of 156 LPD surgical videos from 4 centers were included in this study for analysis. The core benchmark average values derived using the ABC method were as follows: the total LPD surgical duration was 3 hours; the durations for key surgical steps were 12 minutes for uncinate process dissection, 8.5 minutes for choledochojejunostomy, 18.5 minutes for pancreaticojejunostomy, and 17 minutes for gastrojejunostomy/ duodenojejunostomy. Additionally, the benchmarking evaluated the number of intraoperative bleeding events, the surgical instrument usages, the clippings and scissor cutting events was 14, 19, 29, and 2, respectively.

Conclusions: This study established a multi-dimensional surgical benchmark system for LPD for the first time. This achievement offers pancreatic surgeons objective, data-driven evaluation standards for surgical quality. While helping to accurately locate deficiencies in surgeries clinically, it can effectively optimize the practicality of feedback mechanisms, laying a foundation for enhancing the standardization of surgical operations in pancreatic surgery.