Deep Learning-Based Chest X-Ray Analysis for Multi-Class Lung Disease Detection

Artificial intelligence is evolving very rapidly in different fields, healthcare is one of them. One of the major challenges in healthcare is the diagnosis of lung diseases, such as lung cancer, pneumonia, and tuberculosis, where accurate clarification of chest X-rays is very important. Traditional Deep Learning (DL) models like YOLOv12, optimized mostly for object detection and ResNet101, Deep CNN proved to have limitations such as low recall for some disease classes and poor generalization to unseen data. To mitigate these limitations, this research proposes an architecture called ResAttn-Net that uses Squeeze-And-Excitation (SE) blocks to enable channel-wise attention and Multi-Scale Feature Fusion (MSFF) to extract multi-resolution features. Based on an adapted ResNet34 backbone, this model is designed particularly for improved feature extraction in chest X-ray images. Extreme testing on a large chest X-ray dataset has shown that proposed model obtains a validation accuracy of 98.5% and a test accuracy of 96.5%, outperforming YOLOv12 (with 95% validation accuracy and 91% test accuracy) and ResNet101 (with 93% validation accuracy and 90% test accuracy). These results highlight the strength of Proposed model as a computationally light yet strong AI-powered lung disease classification tool.