Deep-learning system for real-time differentiation between Crohn's disease, intestinal Behçet's disease, and...

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Differentiating diseases with similar endoscopic findings, such as intestinal Behçet's disease (BD), Crohn's disease (CD), and intestinal tuberculosis (ITB), is clinically challenging.
• Big data pattern analysis offers potential for improved diagnostic accuracy in gastroenterology.

Purpose of the Study:

• To develop and evaluate a deep-learning algorithm for the differential diagnosis of intestinal BD, CD, and ITB using colonoscopy images.

Main Methods:

• A convolutional neural network (CNN) model was implemented using Pytorch.
• Gradient-weighted Class Activation Mapping was used for model visualization.
• Algorithm performance was assessed using the area under the receiver operating characteristic curve (AUROC).

Main Results:

• The study analyzed 6617 colonoscopy images from 211 CD, 299 intestinal BD, and 217 ITB patients.
• The CNN demonstrated superior discrimination for typical disease images compared to all images (e.g., 72.01% vs. 65.15% for all three diseases).
• AUROC values for typical images ranged from 0.8211 to 0.9360, indicating strong diagnostic performance.

Conclusions:

• A deep-learning model can effectively discriminate between colonoscopy images of intestinal BD, CD, and ITB.
• The algorithm showed enhanced diagnostic accuracy when utilizing typical disease images.
• This AI-driven approach offers a valuable tool for the differential diagnosis of these challenging gastrointestinal conditions.