A Robust Deep Model for Classification of Peptic Ulcer and Other Digestive Tract Disorders Using Endoscopic Images.

Area of Science:

• Biomedical imaging
• Artificial intelligence in healthcare
• Gastroenterology

Background:

• Gastrointestinal (GI) diseases like peptic ulcers and stomach cancer are common.
• Wireless Capsule Endoscopy (WCE) offers painless GI imaging but generates vast datasets.
• Existing deep learning models struggle with accurate and efficient WCE image analysis.

Purpose of the Study:

• To develop an efficient and accurate deep learning model for classifying GI tract abnormalities from WCE images.
• To introduce a custom Convolutional Neural Network (CNN) architecture, GIDD-Net, for improved WCE image analysis.
• To address the class imbalance issue in WCE datasets for robust model performance.

Main Methods:

• A custom CNN architecture, GI Disease-Detection Network (GIDD-Net), was designed from scratch.
• The Borderline Synthetic Minority Over-sampling Technique (BL-SMOTE) was employed to handle class imbalance.
• Class activation patterns were visualized as heatmaps for disorder localization.

Main Results:

• The GIDD-Net model achieved high performance across key metrics: 98.9% accuracy, 99.8% AUC, 98.9% F1-score, 98.9% precision, and 98.8% recall.
• The model demonstrated superior performance compared to other state-of-the-art models.
• Low loss value of 0.0474 was recorded, indicating model efficiency.

Conclusions:

• The proposed GIDD-Net offers a reliable and efficient solution for GI disease detection using WCE images.
• The model's low computational cost and high accuracy make it suitable for real-time analysis.
• GIDD-Net shows significant potential to improve the diagnosis and management of GI disorders.