Scholarlink

<p>Laparoscopic cholecystectomy is the gold standard for treating symptomatic gallstones. Despite established safety strategies, variability in their adoption leads to inconsistent outcomes. Artificial intelligence (AI) offers potential for real-time recognition of procedural stages, contributing to surgical standardization and safety. A prospective observational study was conducted between February 2023 and March 2024. Patients undergoing laparoscopic cholecystectomy for symptomatic gallstones or class 1 cholecystitis (Tokyo classification) were included. From 141 surgical videos, 131 were used to extract 1,841 annotated images representing seven predefined stages: cavity entry, port placement, gallbladder traction, dissection, clipping and sectioning, specimen removal, and final checking. A ResNet-50 neural network was trained and validated using a 90/10 split. Performance was evaluated on 10 unseen videos and tested live in 10 surgeries, comparing AI stage detection with expert surgical judgment. Validation accuracy ranged from 80.15 to 91.74%. In live testing, the system achieved 100% agreement with expert surgeons. The dissection phase was consistently the longest. Real-time processing speed ranged from 1 to 2 frames per second. AI-based recognition of laparoscopic cholecystectomy stages is feasible and demonstrates high concordance with expert evaluation. This approach may support procedural standardization and improve intraoperative decision-making. Further multicenter validation is recommended.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Artificial intelligence in the recognition of stages in live laparoscopic cholecystectomy

  • 期刊论文

  • 发表日期：

    2026年2月19日

查看全文
  • Sean Olivieri,
  • Enrique Petracchi,
  • Hugo Zandalazini

摘要

Laparoscopic cholecystectomy is the gold standard for treating symptomatic gallstones. Despite established safety strategies, variability in their adoption leads to inconsistent outcomes. Artificial intelligence (AI) offers potential for real-time recognition of procedural stages, contributing to surgical standardization and safety. A prospective observational study was conducted between February 2023 and March 2024. Patients undergoing laparoscopic cholecystectomy for symptomatic gallstones or class 1 cholecystitis (Tokyo classification) were included. From 141 surgical videos, 131 were used to extract 1,841 annotated images representing seven predefined stages: cavity entry, port placement, gallbladder traction, dissection, clipping and sectioning, specimen removal, and final checking. A ResNet-50 neural network was trained and validated using a 90/10 split. Performance was evaluated on 10 unseen videos and tested live in 10 surgeries, comparing AI stage detection with expert surgical judgment. Validation accuracy ranged from 80.15 to 91.74%. In live testing, the system achieved 100% agreement with expert surgeons. The dissection phase was consistently the longest. Real-time processing speed ranged from 1 to 2 frames per second. AI-based recognition of laparoscopic cholecystectomy stages is feasible and demonstrates high concordance with expert evaluation. This approach may support procedural standardization and improve intraoperative decision-making. Further multicenter validation is recommended.