Related Experiment Video
Updated: May 28, 2025
11:03
Robotic Left Hepatectomy using Indocyanine Green Fluorescence Imaging for an Intrahepatic Complex Biliary Cyst
Published on: June 24, 2022
5.5K
Artificial intelligence for automatic diagnosis and pleomorphic morphological characterization of malignant biliary
Miguel Mascarenhas1,2,3,4, Maria João Almeida5,6, Mariano González-Haba7
1Department of Gastroenterology, Precision Medicine Unit, São João University Hospital, Porto, Portugal. miguelmascarenhassaraiva@gmail.com.
Scientific Reports
|February 14, 2025
Summary
Artificial intelligence (AI) using convolutional neural networks (CNNs) can accurately detect malignant biliary strictures from digital single-operator cholangioscopy (D-SOC) images. This AI tool shows promise for improving the diagnosis of biliary tract cancers.
Area of Science:
- Gastroenterology and Hepatology
- Medical Imaging and Artificial Intelligence
- Oncology
Background:
- Diagnosing biliary strictures (BS) is challenging, often requiring invasive procedures.
- Digital single-operator cholangioscopy (D-SOC) provides visual data for BS characterization.
- Artificial intelligence (AI) offers potential for automated analysis of medical images.
Purpose of the Study:
- To validate a convolutional neural network (CNN) model for detecting and characterizing malignant biliary strictures using D-SOC images.
- To assess the performance of AI in differentiating malignant from benign biliary strictures.
- To contribute a diverse, multicenter dataset for AI research in biliary tract diseases.
Main Methods:
- A multicenter dataset of 96,020 D-SOC images from 164 patients was retrospectively analyzed.
- A CNN model was trained on 90% of the data and validated on 10%, focusing on features like tumor vessels, papillary projections, nodules, and masses.
- Performance was evaluated using metrics including accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the receiver operating characteristic curve (AUROC).
Main Results:
- The CNN model achieved high diagnostic performance: 92.9% accuracy, 91.7% sensitivity, 94.4% specificity, 95.1% PPV, 93.1% NPV, and an AUROC of 0.95.
- Accuracy for specific morphological features varied: papillary projections (90.8%), nodules (93.6%), masses (93.2%), and tumor vessels (78.1%).
- The study analyzed a large dataset comprising 50,427 malignant stricture images and 45,593 benign findings.
Conclusions:
- AI-driven CNN models demonstrate significant potential to enhance diagnostic accuracy for suspected malignant biliary strictures.
- The developed AI tool can effectively analyze D-SOC images for automated detection and characterization of biliary malignancies.
- This multicenter validation supports the broader application of AI in diagnosing challenging biliary tract conditions.
