DeepPulmoAI: Hybrid Neural Network Selection for Automated Pulmonary Disease Classification

Abstract

Pulmonary diseases such as pneumonia, tuberculosis, lung cancer, and chronic obstructive pulmonary disease (COPD) represent major global health challenges and require accurate early diagnosis for effective treatment. Medical imaging techniques including chest X-rays and computed tomography (CT) scans are widely used for pulmonary disease detection. However, manual interpretation of pulmonary images is time-consuming and highly dependent on radiological expertise. To address these limitations, this research proposes PulmoNet-X, an intelligent ensemble deep learning framework for pulmonary image classification using appropriate neural network selection and ensemble learning techniques. The proposed system integrates multiple deep learning architectures, including Convolutional Neural Networks (CNN), DenseNet, ResNet, and EfficientNet models, to automatically extract discriminative features from pulmonary images. An adaptive neural network selection mechanism identifies the most suitable model based on image characteristics and classification performance. Ensemble learning strategies are further employed to combine the outputs of multiple neural networks, thereby improving classification accuracy, robustness, and generalization capability. The framework incorporates preprocessing techniques such as image normalization, noise reduction, augmentation, and segmentation to enhance image quality and improve feature extraction efficiency. Explainable Artificial Intelligence (XAI) methods such as Grad-CAM and SHAP are integrated to visualize affected pulmonary regions and provide interpretable diagnostic insights for healthcare professionals. Experimental evaluation is performed using benchmark pulmonary imaging datasets, and the proposed model is compared with existing machine learning and deep learning approaches. Results demonstrate superior classification accuracy, precision, recall, F1-score, and reduced false detection rates. The proposed PulmoNet-X system provides a reliable, scalable, and explainable solution for automated pulmonary disease diagnosis and supports clinicians in making accurate and timely medical decisions.