INTELCAPE: A Deep Learning-Powered System for Automated, High-Accuracy Crohn's Disease Diagnosis via Capsule...

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Capsule endoscopy (CE) is crucial for diagnosing Crohn's disease (CD).
• Manual interpretation of CE videos is time-consuming and prone to errors.
• Automating CE analysis can improve diagnostic accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate INTELCAPE, an AI system for automated CE video analysis.
• To enhance the accuracy and efficiency of Crohn's disease diagnosis using CE.
• To benchmark INTELCAPE's performance against human clinicians.

Main Methods:

• A multi-task deep learning framework was employed for small-intestine segmentation, lesion detection, and CD diagnosis.
• The study utilized CE videos from 757 (Cohort 1) and 115 (Cohort 2) patients across two Chinese hospitals.
• INTELCAPE integrated ResNet, Transformer, and EfficientNet architectures for hierarchical video processing.

Main Results:

• INTELCAPE achieved high performance in segmentation (IoU > 94%), lesion detection (AUC > 0.980), and CD diagnosis (AUC > 0.982).
• Diagnostic accuracy for CD was 90%, comparable to specialists (93.33%) but significantly faster (10-fold reduction in interpretation time).
• The AI system improved overall diagnostic accuracy (76.7% to 94.8%) and reduced interpretation time (67.9 to 22.5 min).

Conclusions:

• INTELCAPE effectively automates CE video analysis for Crohn's disease diagnosis.
• The AI system enhances diagnostic accuracy and efficiency, particularly benefiting less experienced clinicians.
• INTELCAPE represents a significant advancement in AI-assisted gastroenterological diagnostics.