Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Efficient and Accurate Pneumonia Detection Using a Novel Multi-Scale Transformer Approach

Alireza Saber, Amirreza Fateh, Pouria Parhami, Alimohammad Siahkarzadeh, Mansoor Fateh, Saideh Ferdowsi

TL;DR

This work addresses automated pneumonia detection from chest X-rays by integrating precise lung segmentation with efficient multi-scale feature processing. It combines a lightweight TransUNet-based segmentation module with a ResNet-inspired backbone and a CRAM-attentive transformer to fuse multi-scale information for classification. The approach achieves Dice scores of $95.68$ on segmentation and classification accuracies of $93.75 ext{\%}$ (Kermany) and $96.04 ext{\%}$ (Cohen), using only $2.29$ million learnable parameters, demonstrating strong performance with low computational cost. The method is well-suited for resource-constrained clinical environments and offers explainability via Grad-CAM, indicating practical potential for global pneumonia screening.

Abstract

Pneumonia, a prevalent respiratory infection, remains a leading cause of morbidity and mortality worldwide, particularly among vulnerable populations. Chest X-rays serve as a primary tool for pneumonia detection; however, variations in imaging conditions and subtle visual indicators complicate consistent interpretation. Automated tools can enhance traditional methods by improving diagnostic reliability and supporting clinical decision-making. In this study, we propose a novel multi-scale transformer approach for pneumonia detection that integrates lung segmentation and classification into a unified framework. Our method introduces a lightweight transformer-enhanced TransUNet for precise lung segmentation, achieving a Dice score of 95.68% on the "Chest X-ray Masks and Labels" dataset with fewer parameters than traditional transformers. For classification, we employ pre-trained ResNet models (ResNet-50 and ResNet-101) to extract multi-scale feature maps, which are then processed through a modified transformer module to enhance pneumonia detection. This integration of multi-scale feature extraction and lightweight transformer modules ensures robust performance, making our method suitable for resource-constrained clinical environments. Our approach achieves 93.75% accuracy on the "Kermany" dataset and 96.04% accuracy on the "Cohen" dataset, outperforming existing methods while maintaining computational efficiency. This work demonstrates the potential of multi-scale transformer architectures to improve pneumonia diagnosis, offering a scalable and accurate solution to global healthcare challenges. https://github.com/amirrezafateh/Multi-Scale-Transformer-Pneumonia

Efficient and Accurate Pneumonia Detection Using a Novel Multi-Scale Transformer Approach

TL;DR

This work addresses automated pneumonia detection from chest X-rays by integrating precise lung segmentation with efficient multi-scale feature processing. It combines a lightweight TransUNet-based segmentation module with a ResNet-inspired backbone and a CRAM-attentive transformer to fuse multi-scale information for classification. The approach achieves Dice scores of on segmentation and classification accuracies of (Kermany) and (Cohen), using only million learnable parameters, demonstrating strong performance with low computational cost. The method is well-suited for resource-constrained clinical environments and offers explainability via Grad-CAM, indicating practical potential for global pneumonia screening.

Abstract

Pneumonia, a prevalent respiratory infection, remains a leading cause of morbidity and mortality worldwide, particularly among vulnerable populations. Chest X-rays serve as a primary tool for pneumonia detection; however, variations in imaging conditions and subtle visual indicators complicate consistent interpretation. Automated tools can enhance traditional methods by improving diagnostic reliability and supporting clinical decision-making. In this study, we propose a novel multi-scale transformer approach for pneumonia detection that integrates lung segmentation and classification into a unified framework. Our method introduces a lightweight transformer-enhanced TransUNet for precise lung segmentation, achieving a Dice score of 95.68% on the "Chest X-ray Masks and Labels" dataset with fewer parameters than traditional transformers. For classification, we employ pre-trained ResNet models (ResNet-50 and ResNet-101) to extract multi-scale feature maps, which are then processed through a modified transformer module to enhance pneumonia detection. This integration of multi-scale feature extraction and lightweight transformer modules ensures robust performance, making our method suitable for resource-constrained clinical environments. Our approach achieves 93.75% accuracy on the "Kermany" dataset and 96.04% accuracy on the "Cohen" dataset, outperforming existing methods while maintaining computational efficiency. This work demonstrates the potential of multi-scale transformer architectures to improve pneumonia diagnosis, offering a scalable and accurate solution to global healthcare challenges. https://github.com/amirrezafateh/Multi-Scale-Transformer-Pneumonia
Paper Structure (38 sections, 23 equations, 8 figures, 6 tables)

This paper contains 38 sections, 23 equations, 8 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Segmentation
  4. U-Net for CXR Segmentation
  5. U-Net Enhancements with Transformers
  6. Classification
  7. Classical Approaches for CXR Classification
  8. Deep Learning Models
  9. Transfer Learning
  10. Ensemble Approaches
  11. Transformers
  12. Proposed method
  13. Overview
  14. Segmentation task
  15. TransUNet Architecture
  16. ...and 23 more sections

Figures (8)

  • Figure 1: The block diagram of the proposed method
  • Figure 2: The TransUnet Architecture
  • Figure 3: Applying trained TransUNet to predict lung masks on "Cohen"/"Kermay" datasets
  • Figure 4: The overview of the proposed method on classification task
  • Figure 5: The overview of CRAM
  • ...and 3 more figures