Enhanced COVID-19 detection from chest X-rays using ERBMAHE and a channel attention-based hybrid CNN model

Abstract

The Coronavirus disease (COVID-19) has significantly impacted global health, creating an urgent demand for efficient and accurate diagnostic methods. Chest X-ray (CXR) imaging has emerged as an essential tool for early detection; however, issues such as low contrast and poor image quality can hinder accurate classification. To address these challenges, this study proposes a novel hybrid deep learning model that integrates VGG16 and ResNet101 architectures through a channel-wise attention fusion mechanism. This innovative design allows the model to adaptively emphasize the most informative features across both networks, improving diagnostic accuracy. To further enhance performance, we introduce Exposure Region-Based Modified Adaptive Histogram Equalization (ERBMAHE), a novel image enhancement technique that improves contrast and detail visibility in CXR images. Additionally, we employ Gradient-weighted Class Activation Mapping (Grad-CAM) to visualize and interpret the discriminative regions used by the model during prediction, thereby enhancing the model’s explainability and clinical trustworthiness. The system is trained using five-fold cross-validation with data augmentation to ensure robustness and generalization. Experimental results demonstrate that the proposed attention-based hybrid model, when paired with ERBMAHE-enhanced images, achieves a test accuracy of 97.94%, precision of 99.06%, F1-score of 97.92%, and recall of 96.80%, outperforming existing state-of-the-art methods. This framework provides a promising tool for rapid, interpretable, and reliable COVID-19 diagnosis, particularly in resource-limited healthcare settings.