Comparative Analysis of Deep Learning Architectures for Multi-Disease Classification of Single-Label Chest X-rays

Background: Chest X-ray imaging has been the most widely used diagnostic technique for pulmonary and cardiac disorders in healthcare systems around the world, owing to its low cost and ease of use. However, the accuracy of the diagnosis is severely hampered by a lack of radiologists and inter-observer variability, which is exacerbated by resource-constrained circumstances. Even though deep learning methods for disease classification have shown great promise, there has been a lack of rigorous comparative evaluations of contemporary architectures for prediction using balanced multi-class chest disease data. Aims: This work examines seven popular deep learning models for multi-class Chest X-ray classification, with an emphasis on trade-offs between performance indicators and computational efficiency. The findings would inform deployment decisions in healthcare settings with diverse resource availability. Methodology: The architectures studied were ConvNeXt-Tiny, DenseNet121, DenseNet201, ResNet50, Vision Transformer (ViT-B/16), EfficientNetV2-M, and MobileNetV2. A comprehensive dataset was generated from existing repositories, consisting of 13,108 training photos, 1,455 validation images, and 3,517 test images for five conditions: Cardiomegaly, COVID-19, Normal, Pneumonia, and Tuberculosis. All models were initialized with ImageNet-pretrained weights and trained under consistent settings, including standardized preprocessing, data augmentation, and optimization hyperparameters. To evaluate model performance, we used metrics such as AUROC, overall accuracy, precision, recall, F1-score, and computational efficiency. Results: All seven studied designs achieved test accuracies that exceeded 90%. ConvNeXt Tiny performed well, with a validation AUROC of 98.64% and a test accuracy of 92.31%. ResNet50 followed closely with 92% test accuracy, while ViT-B/16 earned 91.87%. Notably, MobileNetV2 emerged as the most parameter-efficient Net alternative, with only 3.50 million parameters. Despite its small size, it achieved a test AUROC of 94.10% and obtained the highest efficiency rating in our study. This lightweight architecture achieved around 98.3% of the accuracy of the best-performing model while using 87.5% fewer parameters, which has important implications for deployment in resource-constrained contexts. Conclusion: The current findings show that excellent accuracy in multi-disease categorization of Chest X-ray pictures is possible without requiring significant computational resources. This finding has important implications for the practical integration of deep learning as a diagnostic aid in a variety of healthcare settings, both resource-rich and resource-constrained. Furthermore, choosing a suitable architecture should consider the available infrastructure as well as the unique characteristics of each deployment scenario.