Interpreting chest radiographs is a challenging and resource-intensive clinical task, largely because of the inherent complexity of identifying a broad spectrum of pulmonary pathologies. As a result, there is a pressing need for innovative techniques that can accurately classify multiple abnormalities in chest X-ray images. This research presents an enhanced deep learning framework tailored for multi-label classification of chest X-ray images, addressing a comprehensive set of conditions, including lung opacity, normal pulmonary findings, COVID-19, bacterial pneumonia, viral pneumonia, and tuberculosis. The proposed approach utilizes a refined version of the VGG16 architecture augmented with squeeze-and-excitation (SE) blocks. This model was trained on an extensive collection of chest X-ray images and rigorously compared with leading contemporary methods using key performance indicators: accuracy, F1-score, precision, recall, and area under the curve (AUC). The modified VGG16-SE model outperformed existing techniques across every evaluation metric. It recorded an accuracy of 98.49%, an F1-score of 98.23%, a precision of 98.41%, a recall of 98.07%, and an AUC of 98.86%. The current study offers a highly effective deep learning strategy for classifying chest radiographs. Given its strong performance across diverse lung conditions, the model shows considerable promise for seamless incorporation into routine clinical practice, supporting faster and more precise diagnosis of pulmonary diseases.
