Comparative Analysis of Deep Learning Models for COVID-19 Detection in X-Ray Images

The COVID-19 pandemic has necessitated the development of rapid, accurate, and automated diagnostic tools to assist healthcare professionals in patient screening and management. This comprehensive study presents a detailed comparative analysis of ten state-of-the-art Convolutional Neural Network (CNN) architectures for COVID-19 detection using chest X-ray radiography images. The research systematically evaluates DenseNet, ConvNeXTiny, EfficientNetB0, ResNet50, VGG19, VGG16, MobileNet, GoogleNet, AlexNet, and LeNet models using the COVID-19 Radiography Database. Each model was rigorously trained and validated using standardized protocols to ensure fair comparison. Comprehensive performance metrics including validation accuracy, test accuracy, loss functions, convergence analysis, and overfitting assessment were employed. The experimental results demonstrate that modern architectures, particularly DenseNet and ConvNeXTiny, achieved superior performance with 97% validation accuracy and 96% test accuracy, exhibiting excellent generalization capabilities with minimal overfitting (1.00% accuracy difference). The study also analyzes computational efficiency, training dynamics, and practical deployment considerations. These findings provide crucial insights for developing robust computer-aided diagnosis systems for COVID-19 detection and establish benchmarks for future research in medical image classification.