Diagnosing COVID-19 Severity from Chest X-Ray Images Using ViT and CNN Architectures

2025-02-23

AI Summary

This study addresses the quantitative assessment of COVID-19 severity from chest X-ray (CXR) images. We introduce the first large-scale, multi-source severity-annotated dataset supporting both regression and three-class classification tasks, and conduct the first systematic comparison of Vision Transformers (ViT) and convolutional neural networks (CNNs) for this purpose. Methodologically, we propose a novel continuous severity regression paradigm grounded in radiologist-derived scores—overcoming limitations of discrete grading—and employ DenseNet-161 and ViT with ImageNet and CXR-specific transfer learning strategies. Results show that DenseNet-161 achieves 80.0% accuracy in three-class classification, while the ViT-based regression model attains a mean absolute error (MAE) of 0.5676, outperforming existing approaches. To foster reproducibility and clinical interpretability, we publicly release all code and annotation protocols.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Security and Privacy: Large-scale security measurements
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

The COVID-19 pandemic strained healthcare resources and prompted discussion about how machine learning can alleviate physician burdens and contribute to diagnosis. Chest x-rays (CXRs) are used for diagnosis of COVID-19, but few studies predict the severity of a patient's condition from CXRs. In this study, we produce a large COVID severity dataset by merging three sources and investigate the efficacy of transfer learning using ImageNet- and CXR-pretrained models and vision transformers (ViTs) in both severity regression and classification tasks. A pretrained DenseNet161 model performed the best on the three class severity prediction problem, reaching 80% accuracy overall and 77.3%, 83.9%, and 70% on mild, moderate and severe cases, respectively. The ViT had the best regression results, with a mean absolute error of 0.5676 compared to radiologist-predicted severity scores. The project's source code is publicly available.

Problem

• Diagnose COVID-19 severity using chest X-rays
• Evaluate ViT and CNN architectures for severity prediction
• Assess transfer learning efficacy on COVID severity dataset

Innovation

• ViT for COVID severity regression
• DenseNet161 for severity classification
• Transfer learning from CXR and ImageNet