High-precision classification of WCE-based gastrointestinal abnormality using a fusion deep learning approach.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Gastroenterology

Background:

• Gastrointestinal abnormalities are a global health concern, with early detection significantly reducing mortality.
• Endoscopy, particularly Wireless Capsule Endoscopy (WCE), is a vital, minimally invasive diagnostic tool for the gastrointestinal tract.
• Accurate classification of WCE findings is essential for timely diagnosis and treatment.

Purpose of the Study:

• To enhance the early detection of specific gastrointestinal disorders using Wireless Capsule Endoscopy (WCE).
• To develop and evaluate a novel deep learning model for improved classification of WCE images.
• To specifically improve the detection of ulcerative colitis, polyps, and dyed-lifted polyps.

Main Methods:

• Proposed a fusion deep learning (DL) model integrating Convolutional Neural Networks (CNNs) and Graph Neural Networks (GNNs).
• CNNs were utilized for extracting relational information from image regions.
• GNNs were employed to analyze spatial structure traits within WCE images.
• The model was trained and validated on publicly available WCE datasets.

Main Results:

• The proposed fusion DL model achieved a high accuracy of 98.82% for classifying WCE findings.
• The model demonstrated superior performance compared to traditional CNN architectures.
• Outperformed previously developed pre-trained models in WCE disorder classification.

Conclusions:

• The fusion DL model offers a significant advancement in WCE-based gastrointestinal disorder detection.
• This approach shows promise for more accurate and efficient diagnosis of ulcerative colitis, polyps, and dyed-lifted polyps.
• The integration of CNNs and GNNs provides a robust framework for analyzing complex medical image data.