Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CheX-DS: Improving Chest X-ray Image Classification with Ensemble Learning Based on DenseNet and Swin Transformer

Xinran Li, Yu Liu, Xiujuan Xu, Xiaowei Zhao

TL;DR

CheX-DS tackles long-tail, multi-label chest X-ray classification by fusing DenseNet121 and Swin Transformer in an ensemble, leveraging both local and global feature representations. It introduces a loss that combines weighted binary cross-entropy with an asymmetric term to balance inter- and intra-class imbalances, and optimizes ensemble weights via differential evolution. On the NIH ChestX-ray14 dataset, it achieves an average AUROC of $83.76\%$, outperforming several state-of-the-art baselines. This approach demonstrates the value of integrating CNN and Transformer features with a data-tailored loss for improved automated chest X-ray diagnosis in imbalanced settings.

Abstract

The automatic diagnosis of chest diseases is a popular and challenging task. Most current methods are based on convolutional neural networks (CNNs), which focus on local features while neglecting global features. Recently, self-attention mechanisms have been introduced into the field of computer vision, demonstrating superior performance. Therefore, this paper proposes an effective model, CheX-DS, for classifying long-tail multi-label data in the medical field of chest X-rays. The model is based on the excellent CNN model DenseNet for medical imaging and the newly popular Swin Transformer model, utilizing ensemble deep learning techniques to combine the two models and leverage the advantages of both CNNs and Transformers. The loss function of CheX-DS combines weighted binary cross-entropy loss with asymmetric loss, effectively addressing the issue of data imbalance. The NIH ChestX-ray14 dataset is selected to evaluate the model's effectiveness. The model outperforms previous studies with an excellent average AUC score of 83.76\%, demonstrating its superior performance.

CheX-DS: Improving Chest X-ray Image Classification with Ensemble Learning Based on DenseNet and Swin Transformer

TL;DR

CheX-DS tackles long-tail, multi-label chest X-ray classification by fusing DenseNet121 and Swin Transformer in an ensemble, leveraging both local and global feature representations. It introduces a loss that combines weighted binary cross-entropy with an asymmetric term to balance inter- and intra-class imbalances, and optimizes ensemble weights via differential evolution. On the NIH ChestX-ray14 dataset, it achieves an average AUROC of , outperforming several state-of-the-art baselines. This approach demonstrates the value of integrating CNN and Transformer features with a data-tailored loss for improved automated chest X-ray diagnosis in imbalanced settings.

Abstract

The automatic diagnosis of chest diseases is a popular and challenging task. Most current methods are based on convolutional neural networks (CNNs), which focus on local features while neglecting global features. Recently, self-attention mechanisms have been introduced into the field of computer vision, demonstrating superior performance. Therefore, this paper proposes an effective model, CheX-DS, for classifying long-tail multi-label data in the medical field of chest X-rays. The model is based on the excellent CNN model DenseNet for medical imaging and the newly popular Swin Transformer model, utilizing ensemble deep learning techniques to combine the two models and leverage the advantages of both CNNs and Transformers. The loss function of CheX-DS combines weighted binary cross-entropy loss with asymmetric loss, effectively addressing the issue of data imbalance. The NIH ChestX-ray14 dataset is selected to evaluate the model's effectiveness. The model outperforms previous studies with an excellent average AUC score of 83.76\%, demonstrating its superior performance.
Paper Structure (17 sections, 3 equations, 7 figures, 3 tables)

This paper contains 17 sections, 3 equations, 7 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Convolutional Neural Networks in medical image domain
  4. Transformer in medical image domain
  5. Method
  6. DenseNet
  7. Swin Transformer
  8. Loss Function
  9. Ensemble Learning
  10. Experiments
  11. Dataset
  12. Experimental Setup
  13. Evaluation Metrics
  14. Ensemble Model CheX-DS vs. Individual Models
  15. Comparison of loss functions
  16. ...and 2 more sections

Figures (7)

  • Figure 1: The architecture diagram of DenseNet121
  • Figure 2: The schematic diagram of CheX-DS, featuring the average weighted ensemble with differential evolution.
  • Figure 3: The architecture diagram of Swin Transformer
  • Figure 4: The distribution of the number of cases for the 14 diseases
  • Figure 5: ROC Curves of improved loss DenseNet with 15 classes
  • ...and 2 more figures