Enhanced COVID-19 Detection Using Attention-Augmented ResNet with Inception Modules and Regularization Techniques

This research proposes a novel deep-learning model for accurate COVID-19 detection from lung CT scans. By enhancing the ResNet-50 architecture with Convolutional Block Attention Modules (CBAM) and Inception modules, the model effectively captures multi-scale features and refines attention, improving diagnostic accuracy. A comprehensive preprocessing pipeline, including data augmentation, ensures the model’s generalizability across diverse datasets. The model was validated using using publicly available balanced datasets and a 5 -fold cross-validation strategy, demonstrating superior performance compared to traditional models like DenseNet and EfficientNet. The proposed approach achieved a highest notable ${9 9. 8 1 \%}$ accuracy in COVID-19 classification, outperforming existing methods. These results highlight the model’s potential as a reliable diagnostic tool in clinical settings, particularly for early detection and treatment planning. Future work will focus on expanding the dataset and exploring ensemble methods to further improve generalizability.