Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SynthEnsemble: A Fusion of CNN, Vision Transformer, and Hybrid Models for Multi-Label Chest X-Ray Classification

S. M. Nabil Ashraf, Md. Adyelullahil Mamun, Hasnat Md. Abdullah, Md. Golam Rabiul Alam

TL;DR

This study tackles multi-label thoracic disease classification from chest X-rays using the ChestX-ray14 dataset. It systematically compares CNNs, vision transformers, and CNN–transformer hybrids, and introduces a differential-evolution–driven weighted ensemble to boost accuracy. The best single model, CoAtNet, achieves AUROC 0.842, while the ensemble reaches 0.854, surpassing state-of-the-art methods. The results demonstrate the value of ensemble learning that leverages diverse architectures for robust medical imaging diagnosis with practical impact for automated thoracic disease screening.

Abstract

Chest X-rays are widely used to diagnose thoracic diseases, but the lack of detailed information about these abnormalities makes it challenging to develop accurate automated diagnosis systems, which is crucial for early detection and effective treatment. To address this challenge, we employed deep learning techniques to identify patterns in chest X-rays that correspond to different diseases. We conducted experiments on the "ChestX-ray14" dataset using various pre-trained CNNs, transformers, hybrid(CNN+Transformer) models and classical models. The best individual model was the CoAtNet, which achieved an area under the receiver operating characteristic curve (AUROC) of 84.2%. By combining the predictions of all trained models using a weighted average ensemble where the weight of each model was determined using differential evolution, we further improved the AUROC to 85.4%, outperforming other state-of-the-art methods in this field. Our findings demonstrate the potential of deep learning techniques, particularly ensemble deep learning, for improving the accuracy of automatic diagnosis of thoracic diseases from chest X-rays. Code available at:https://github.com/syednabilashraf/SynthEnsemble

SynthEnsemble: A Fusion of CNN, Vision Transformer, and Hybrid Models for Multi-Label Chest X-Ray Classification

TL;DR

This study tackles multi-label thoracic disease classification from chest X-rays using the ChestX-ray14 dataset. It systematically compares CNNs, vision transformers, and CNN–transformer hybrids, and introduces a differential-evolution–driven weighted ensemble to boost accuracy. The best single model, CoAtNet, achieves AUROC 0.842, while the ensemble reaches 0.854, surpassing state-of-the-art methods. The results demonstrate the value of ensemble learning that leverages diverse architectures for robust medical imaging diagnosis with practical impact for automated thoracic disease screening.

Abstract

Chest X-rays are widely used to diagnose thoracic diseases, but the lack of detailed information about these abnormalities makes it challenging to develop accurate automated diagnosis systems, which is crucial for early detection and effective treatment. To address this challenge, we employed deep learning techniques to identify patterns in chest X-rays that correspond to different diseases. We conducted experiments on the "ChestX-ray14" dataset using various pre-trained CNNs, transformers, hybrid(CNN+Transformer) models and classical models. The best individual model was the CoAtNet, which achieved an area under the receiver operating characteristic curve (AUROC) of 84.2%. By combining the predictions of all trained models using a weighted average ensemble where the weight of each model was determined using differential evolution, we further improved the AUROC to 85.4%, outperforming other state-of-the-art methods in this field. Our findings demonstrate the potential of deep learning techniques, particularly ensemble deep learning, for improving the accuracy of automatic diagnosis of thoracic diseases from chest X-rays. Code available at:https://github.com/syednabilashraf/SynthEnsemble
Paper Structure (21 sections, 2 figures, 6 tables)

This paper contains 21 sections, 2 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Convolutional Neural Networks in medical image domain
  4. Transformer in the medical computer vision
  5. Methodology
  6. Dataset Description
  7. Model Exploration
  8. Hybrid architectures
  9. Vision Transformer (ViT)
  10. Convolutional Neural Network (CNN)
  11. Data pre-processing and splitting
  12. Image Resizing and Normalization
  13. Horizontal Flips and Rotation
  14. Splitting
  15. Training and Optimization
  16. ...and 6 more sections

Figures (2)

  • Figure 1: Disease distribution in the dataset.
  • Figure 4: ROC curve for all 14 diseases displaying the True Positive against False Positive rate, which illustrates the model’s capacity to differentiate effectively.