An Explainable Attention-Guided Quantum Vision Transformer for Lung Disease Classification using Chest X-Ray Images

Abstract

Lung diseases remain a severe world health problem, which contributes to mortality worldwide. Early and accurate diagnosis of pulmonary abnormalities using chest radiographs is essential to improve clinical management and patient outcomes. Although deep learning (DL) models have achieved considerable success in automated lung disease classification, convolutional networks involve challenges in capturing local pathological patterns and global contextual dependencies. Furthermore, growing interest in quantum machine learning has provided new paths for improving feature representation and classification efficiency. To overcome these challenges, the present study proposes a Quantum-Enhanced Vision Transformer Network (QEViT) for automated classification of lung disease from chest X-ray images (CXR). The proposed work integrates EfficientNetB4 for hierarchical feature extraction, Convolutional Block Attention Module (CBAM) for attention-guided feature refinement, Vision Transformer (ViT) for global contextual modelling, and a Variational Quantum Circuit (VQC) for quantum feature generation. The effectiveness of the proposed QEViT was evaluated on the Lung Disease Dataset and the COVID-19 Radiography Database, achieving classification accuracies of 99.8% and 99.7%, respectively. Experimental outcomes proved that integrating DL, attention mechanisms, transformer-based contextual learning, and quantum feature generation improves classification performance. Finally, explainable artificial intelligence (XAI) using Grad-CAM and Score-CAM was incorporated to provide visual interpretations of the model’s predictions. The proposed QEViT network may support automated CXR image analysis for lung disease classification.