Related Experiment Video
Updated: Aug 15, 2026
11:03
Robotic Left Hepatectomy using Indocyanine Green Fluorescence Imaging for an Intrahepatic Complex Biliary Cyst
Published on: June 24, 2022
Artificial intelligence for predicting surgical difficulty in laparoscopic cholecystectomy: a systematic review and
Haoyang Liu1, Xuekai Wu1, Fang Luo2
1The First Clinical College of Chongqing Medical University, Chongqing Medical University, Chongqing, 400016, China.
Surgical Endoscopy
|August 14, 2026
Summary
Predicting laparoscopic cholecystectomy difficulty with artificial intelligence (AI) shows promise, but current models have significant flaws. Further research and validation are needed before clinical use.
Area of Science:
- Surgical Innovation
- Artificial Intelligence in Medicine
- Predictive Analytics
Background:
- Accurate prediction of operative difficulty in laparoscopic cholecystectomy (LC) is crucial for personalized surgical planning and patient safety.
- Current Artificial Intelligence (AI) models for predicting LC difficulty lack substantiated reliability, generalizability, and clinical utility.
- This review critically evaluates the performance and methodological quality of AI models for predicting LC surgical difficulty.
Purpose of the Study:
- To systematically review and evaluate the predictive performance of AI models designed for predicting surgical difficulty in laparoscopic cholecystectomy (LC).
- To assess the methodological quality and risk of bias in existing AI models for LC difficulty prediction.
- To identify promising AI architectures and data integration strategies for future development.
Main Methods:
- A comprehensive literature search was conducted across PubMed, Embase, Web of Science, and Cochrane Library up to March 2, 2026.
- The Prediction Model Risk of Bias Assessment Tool and GRADE framework were employed to assess study quality and evidence certainty.
- Random-effects meta-analysis was used to pool the areas under the curve (AUC) for model performance, adhering to PRISMA guidelines.
Main Results:
- Eighteen studies were included, with sixteen identified as having a high risk of bias.
- Pooled AUC for training and validation models were 0.848 and 0.818, respectively.
- Ensemble models demonstrated superior performance (pooled AUCs of 0.889 and 0.861), and multimodal data integration (clinical, imaging, video) also yielded better results.
Conclusions:
- Existing AI research for predicting LC difficulty is hampered by significant methodological flaws and a high risk of bias.
- External validation is rarely performed, limiting the clinical translation of these predictive models.
- AI models utilizing ensemble architectures and multimodal data warrant further investigation, but require rigorous methodological evaluation and prospective multicenter testing before clinical implementation.
