Artificial Intelligence via Convolutional Neural Network versus Digital Single-Operator Cholangioscopy for the Detection of Malignant Biliary Strictures: A Meta-Analysis

Aims Malignant biliary strictures result from primary (cholangiocarcinoma) or secondary neoplasia with biliary tract extension. The diagnosis of biliary strictures is quite challenging due to limited modalities. Digital Single-Operator Cholangioscopy (DSOC) allows direct visualization of the biliary tree and targeted tissue sampling. Despite its utility, the diagnostic efficiency of DSOC remains suboptimal due to its technical complexity and operator dependence. Artificial intelligence (AI) using convolutional neural network (CNN) have shown great potential for enhancing diagnostic yield through translational invariance, a key feature of image recognition. This study aims to compare the diagnostic performance of AI versus DSOC for the detection of malignant biliary strictures [1] [2] [3].

Methods Major electronic databases and grey literature were searched up to February 2024 for randomized controlled trials assessing the detection rate of malignant biliary strictures between AI and DSOC.

Results The resulting pooled odds ratio of 1.58 (95% CI 1.05 to 2.36) favors Artificial Intelligence through CNN, and showed significant difference in comparison to DSOC (Z=2.21, p=0.03) in detecting malignant biliary strictures. The diamond market also favors the left side and does not intersect the 1 axis, suggesting that the odds ratio is significant. The resulting I2 of 0% (p=0.73) implies that heterogeneity does not exist.

Conclusions Artificial Intelligence-Assisted Cholangioscopy shows significant promise as a new diagnostic tool for identifying malignant biliary strictures, ultimately leading to improved patient outcomes.

Publication History

Article published online:
27 March 2025

• References

• 1 Kwon Y.. et al. 'Evaluating the Efficacy of AI-Enhanced Imaging Techniques in Biliary Stricture Detection Compared to Conventional Methods.'. Journal of Hepatology 2023; 78 (4): 729-738
• 2 Liu Y.. et al. 'Comparison of Deep Learning Techniques and Digital Cholangioscopy for Malignant Biliary Stricture Diagnosis: A Prospective Study.'. Endoscopy International Open 2022; 10 (5): E555-E562
• 3 Wang H.. et al. 'Artificial Intelligence-Based Analysis of Endoscopic Images for the Diagnosis of Biliary Malignancies: A Systematic Review.'. Gastrointestinal Endoscopy 2021; 94 (3): 567-577