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

Gastrointestinal abnormalities are widespread worldwide and pose a significant health challenge. However, their mortality rate can be significantly reduced when they are detected early. Endoscopy is one of the key techniques used to diagnose problems in both the upper and lower gastrointestinal regions. It is also considered less invasive and more patient-friendly than many traditional diagnostic methods. Wireless Capsule Endoscopy (WCE)-based disorder detection and diagnosis have reached a point of convergence, reshaping the landscape. Early diagnosis and appropriate treatment depend on the precise identification and categorization of WCE-based disorders related to medical images. Convolutional neural networks (CNNs) are widely used for disease diagnosis. To improve Wireless Capsule Endoscopy (WCE)-based detection of ulcerative colitis, polyps, dyed-lifted polyps, and normal tissue. We proposed a Fusion deep learning model that combines Graph Neural Networks (GNNs) that analyze spatial structure traits and CNNs that extract relational information from image regions. We used publicly available WCE datasets that assess our models for the classification of Ulcerative colitis, Polyps, and Dyed-lifted polyps. The proposed fusion deep learning (DL) model achieves 98.82% accuracy. The results demonstrate that the suggested model outperforms the traditional CNN architecture and earlier pre-trained models.