Attention-Guided Diagnostic Intelligence: Towards Trustworthy and Explainable Lung Disease Classification

The rapid advancement of medical imaging technology has revolutionized diagnostics; however, manual interpretation remains subjective and time-intensive. This study presents an attention-based deep learning framework, the Mask-Guided Convolutional Neural Network (MG-CNN), designed to enhance both the classification accuracy and explainability of chest X-ray analysis for COVID-19 and pneumonia. Leveraging the COVID-19 Radiology Dataset, the model integrates a segmentation-based attention mechanism that dynamically prioritizes relevant pulmonary regions while suppressing background noise. Experimental results demonstrate a test accuracy of 89.97%. Crucially, the integration of Gradient-weighted Class Activation Mapping (Grad-CAM) validates that the model’s decision-making aligns with clinical pathology, focusing on lung parenchyma rather than irrelevant artifacts. This work addresses the "black box" limitation of traditional deep learning, offering a transparent, trustworthy Clinical Decision Support System (CDSS) for pulmonary medicine.