Pneumonia Classification from Chest Radiographs using Advanced CovXNet Model

Millions of people worldwide suffer from pneumonia, a serious respiratory illness that raises rates of morbidity and death. It is categorized into bacterial and viral pneumonia, each requiring distinct treatments. Conventional diagnostic techniques, like radiologists’ subjective interpretation of chest X-rays, are prone to misdiagnosis, which emphasizes the need for automated solutions to improve efficiency and accuracy. The identification of pneumonia in medical scans has benefited from the implementation of deep learning models particularly CNNs and VGG16, ResNet50 and EfficientNet. Although these models have proven to be highly accurate in binary classification (pneumonia versus normal), their limited generalizability, high computational costs, and overfitting frequently hinder their ability to perform multiclass classification, which limits their clinical adoption. To address these challenges, we propose AdvancedCovXNet, a novel convolutional neural network (CNN) optimized for multiclass pneumonia classification (normal, bacterial, and viral pneumonia). The model integrates dilated atrous convolutions to capture fine-grained spatial features and depthwise separable convolutions to improve computational efficiency. Additionally, advanced data augmentation techniques enhance model robustness, while dropout regularization mitigates overfitting. Our approach leverages deep learning’s ability to extract complex features from medical images, significantly improving classification accuracy and reliability. AdvancedCovXNet serves as a scalable and efficient AI-driven diagnostic tool, offering a promising solution for automated pneumonia detection in clinical settings.