Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Aman Sagar, Preeti Mehta, Monika Shrivastva, Suchi Kumari

TL;DR

The aim is to automate the identification of ten GI abnormality classes, including angioectasia, bleeding, and ulcers, thereby reducing the diagnostic burden on gastroenterologists, and developing an effective AI-driven tool that streamlines the diagnostic process in gastroenterology, ultimately improving patient care and clinical outcomes.

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.

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

TL;DR

The aim is to automate the identification of ten GI abnormality classes, including angioectasia, bleeding, and ulcers, thereby reducing the diagnostic burden on gastroenterologists, and developing an effective AI-driven tool that streamlines the diagnostic process in gastroenterology, ultimately improving patient care and clinical outcomes.

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.

Paper Structure

This paper contains 8 sections, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. Results
  4. Achieved results on the validation dataset
  5. Testing Dataset Information
  6. Discussion
  7. Conclusion
  8. Acknowledgments

Figures (6)

  • Figure 1: Block diagram of the proposed methodology for multi-class classification of gastrointestinal abnormalities in Video Capsule Endoscopy (VCE) frames. The diagram illustrates the key stages, including data collection and preparation, preprocessing, model development with ensemble of DenseNet and ResNet, training and validation evaluations
  • Figure 2: The training and validation loss and accuracy when training dataset split into 80:20 ratio for 50 epochs
  • Figure 3: 2D visualization of features extracted from the ensemble model from the training datset.
  • Figure 4: The confusion matrix results on validation dataset
  • Figure 5: classwise precision, recall, f1-score and overall accuracy of the model on validation dataset
  • ...and 1 more figures