Lung diseases are a leading cause of morbidity and mortality, underscoring the need for accurate diagnostic tools. Chest X-ray imaging is commonly used for diagnosis, but existing deep learning methods often focus on binary classification and struggle with the complexity of lung diseases. To address these challenges, we developed the Multi-Scale Adaptive Attention Fusion Network (MSAAF-Net), a novel framework designed for enhanced lung disease classification. MSAAF-Net integrates multi-scale feature extraction with self-attention, spatial attention, and channel attention mechanisms, dynamically weighted through a class-aware module. This approach enables the model to capture both fine-grained and large-scale pathological features, improving classification across multiple disease classes. Evaluated on a publicly available chest X-ray dataset using five-fold cross-validation, MSAAF-Net achieved a classification accuracy of 93.53%, an F1-score of 93.76%, and an AUC of 98.33%, surpassing state-of-the-art models. These results demonstrate MSAAF-Net’s ability to effectively manage the complexity of multi-class lung disease classification. The framework enhances automated diagnostic accuracy, supporting better clinical decision-making and advancing AI’s role in lung healthcare.