<![CDATA[Tuberculosis and pneumonia continue to pose major global health challenges, particularly in regions with limited radiological resources, where overlapping chest X-ray patterns often complicate differential diagnosis. This study proposes X-RayVision-Net, a hybrid deep learning framework that integrates the Convolutional Block Attention Module (CBAM) into the YOLOv8 architecture to enhance pulmonary disease classification. A quantitative experimental design was employed using 10,056 chest X-ray images categorized as normal, pneumonia, or tuberculosis, collected from multiple public datasets. Image preprocessing involved Contrast Limited Adaptive Histogram Equalization (CLAHE) and balanced data augmentation to improve visual consistency and address class imbalance. The proposed model was trained for 100 epochs and evaluated against a standard YOLOv8 baseline using accuracy, precision, recall, and F1-score. Experimental results demonstrate that the CBAM-enhanced YOLOv8 model achieved an accuracy of 98.99%, outperforming the baseline model (97.37%) and yielding consistent improvements across all performance metrics. The findings confirm that the incorporation of channel and spatial attention mechanisms effectively refines pulmonary feature representation, facilitating more accurate discrimination between tuberculosis and pneumonia. This framework presents a rapid and reliable computer-aided diagnostic approach suitable for deployment in clinical environments with constrained radiology expertise.]]>
Copyrights © 2026
