Gastrointestinal tract disease detection via deep learning based Duo-Feature Optimized Hexa-Classification model

Gastrointestinal tract (GT) diseases impact the entire digestive system, spanning from the mouth to the anus. Wireless Capsule Endoscopy (WCE) stands out as an effective diagnostic tool for Gastrointestinal tract diseases. Nevertheless, accurately identifying various lesion features, such as irregular sizes, shapes, colors, and textures, remains challenging in this field. Several computer vision algorithms have been introduced to tackle these challenges, but many relied on handcrafted features, resulting in inaccuracies in various instances. This research presents a revolutionary Hexa categorization model based on Deep Hexa model has been proposed for the identification the various GT diseases in compressed WCE images. Input images are sourced from the KVASIR and KID datasets, undergoing preprocessing via wavelet-based Retinex and data augmentation. The preprocessed images enter the first phase, where statistical features are extracted using Shuffle network. Simultaneously, the preprocessed images undergo segmentation via UNet. The segmented images then enter the second phase for structural feature extraction. The conditional entropy approach selects Walrus optimization of optimal features from the merged set of structural and statistical features obtained in parallel fusion. Finally, a capsule network is employed to classify GT diseases into hexa categories such as normal, ulcer, pylorus, cecum, esophagitis and polyps based on the selected features. The proposed Deep Hexa Model attains a high accuracy of 99.38 % in GT disease detection. Compared to Duo-deep, ESKNN, VGG-16 Net, Inception-ResNet-v2, and Fusion network, the proposed model improves overall accuracy by 0.25 %, 2.89 %, 3.06 %, 0.90 %, and 1.39 %, respectively.