A real-time interpretable artificial intelligence model for the cholangioscopic diagnosis of malignant biliary stricture (with videos)

Abstract

Background and aims

It is crucial to accurately determine malignant biliary strictures (MBSs) for early curative treatment. The study aimed to develop a real-time interpretable artificial intelligent (AI) system to predict MBSs under digital single-operator cholangioscopy (DSOC).

Methods

A novel interpretable AI system called MBSDeiT was developed, consisting of two models to identify qualified images and then predict MBS in real time. The overall efficiency of MBSDeiT was validated at the image level on internal, external, prospective testing datasets and subgroups analyses, and at the video level on the prospective datasets, and compared with that of endoscopists. The association between AI predictions and endoscopic features was evaluated to increase the interpretability.

Results

MBSDeiT can first automatically select qualified DSOC images with an AUC of 0.904 and 0.921–0.927 on the internal testing dataset and the external testing datasets, and then identify MBSs with an AUC of 0.971 on the internal testing dataset, an AUC of 0.978–0.999 on the external testing datasets, and an AUC of 0.976 on the prospective testing dataset, respectively. MBSDeiT accurately identified 92.3% MBS in prospective testing videos. Subgroups analyses confirmed the stability and robustness of MBSDeiT. MBSDeiT achieved superior performance to that of expert and novice endoscopists. The AI predictions were significantly associated with four endoscopic features (nodular mass; friability; raised intraductal lesion; and abnormal vessels; P < 0.05) under DSOC, which is consistent with the endoscopists' predictions.

Conclusions

The findings suggest that MBSDeiT could be a promising approach for the accurate diagnosis of MBS under DSOC.