Related Experiment Video
Updated: Jul 13, 2025
07:44
Endoscopic Ultrasound-Guided Biliary Drainage: Endoscopic Ultrasound-Guided Hepaticogastrostomy in Malignant Biliary Obstruction
Published on: March 25, 2022
5.8K
Deep Learning for Automatic Diagnosis and Morphologic Characterization of Malignant Biliary Strictures Using Digital
Miguel Mascarenhas Saraiva1,2,3, Tiago Ribeiro1,2, Mariano González-Haba4
1Department of Gastroenterology, São João University Hospital, Alameda Professor Hernâni Monteiro, 4200-427 Porto, Portugal.
Cancers
|October 14, 2023
Summary
Artificial intelligence, using a convolutional neural network (CNN), can now identify malignant biliary strictures (BSs) during digital single-operator cholangioscopy (D-SOC). This AI tool shows high accuracy, potentially improving diagnosis for these critical conditions.
Area of Science:
- Gastroenterology
- Medical Imaging
- Artificial Intelligence
Background:
- Digital single-operator cholangioscopy (D-SOC) aids in diagnosing indeterminate biliary strictures (BSs).
- Pilot studies suggest artificial intelligence (AI) can enhance D-SOC diagnostic capabilities.
Purpose of the Study:
- To develop a convolutional neural network (CNN) for identifying and characterizing malignant BSs using D-SOC images.
- To evaluate the CNN's performance in distinguishing malignant from benign biliary strictures.
Main Methods:
- A CNN was developed and trained on 84,994 D-SOC images from 129 exams across two centers.
- Images were classified as normal/benign or malignant based on histopathology.
- The CNN was assessed for detecting morphological features like tumor vessels and papillary projections.
- Performance metrics included sensitivity, specificity, accuracy, AUROC, and AUPRC.
Main Results:
- The CNN achieved 82.9% overall accuracy, 83.5% sensitivity, and 82.4% specificity.
- The area under the receiver-operating characteristic (AUROC) and precision-recall curves (AUPRC) were 0.92 and 0.93, respectively.
- The model successfully differentiated benign findings from malignant BSs.
Conclusions:
- The developed CNN effectively identifies malignant biliary strictures in D-SOC.
- AI tools integrated with D-SOC hold significant potential to improve diagnostic accuracy for malignant BSs.
