Integrating Deep Feature Extraction and Hybrid ResNet-DenseNet Model for Multi-Class Abnormality Detection in Endoscopic Images

TL;DR

This paper introduces a deep learning ensemble model combining ResNet and DenseNet architectures to accurately classify ten gastrointestinal abnormalities in endoscopic images, aiming to assist clinicians and improve diagnostic efficiency.

Contribution

It presents a novel hybrid ResNet-DenseNet model with robust data preprocessing for multi-class GI abnormality detection, achieving high accuracy and clinical relevance.

Findings

Overall accuracy of 94% on the dataset

Precision varies from 0.56 to 1.00 across classes

Recall reaches 98% for normal findings

Abstract

This paper presents a deep learning framework for the multi-class classification of gastrointestinal abnormalities in Video Capsule Endoscopy (VCE) frames. The aim is to automate the identification of ten GI abnormality classes, including angioectasia, bleeding, and ulcers, thereby reducing the diagnostic burden on gastroenterologists. Utilizing an ensemble of DenseNet and ResNet architectures, the proposed model achieves an overall accuracy of 94\% across a well-structured dataset. Precision scores range from 0.56 for erythema to 1.00 for worms, with recall rates peaking at 98% for normal findings. This study emphasizes the importance of robust data preprocessing techniques, including normalization and augmentation, in enhancing model performance. The contributions of this work lie in developing an effective AI-driven tool that streamlines the diagnostic process in gastroenterology, ultimately improving patient care and clinical outcomes.