Multi-network–Based COVID-19 Detector Using Chest X-ray Images

Abstract

The global outbreak of COVID-19 has emerged as the most catastrophic public health crisis in over five decades, exhibiting profound respiratory implications that primarily compromise pulmonary function. Timely and accurate detection of infection-induced anomalies within the lungs remains critical to curbing the spread and mitigating clinical severity. Among diagnostic imaging modalities, computed tomography (CT) and chest radiography (X-ray) have proven indispensable—offering varying degrees of sensitivity, resolution, and accessibility. To address these limitations, recent advancements in deep learning have introduced data-driven frameworks capable of autonomously learning discriminative features from medical imagery. In this study, we introduce a robust multi-network ensemble architecture tailored for automated multi-class pulmonary infection classification. The proposed framework integrates multiple heterogeneous convolutional neural networks (CNNs), namely, VGG16, VGG19, ResNet50, ResNet101, and EfficientNetB0, each initialized with pre-trained weights derived from large-scale image corpora via transfer learning. Experimental findings indicate that the ensemble model consistently outperforms its constituent networks, attaining a peak classification accuracy of 98.37%. This superior performance underscores the efficacy of multi-network integration in enhancing diagnostic reliability and enabling fine-grained discrimination across complex pathological classes.

Received: 29 October 2024 | Revised: 25 November 2025 | Accepted: 19 January 2026

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

The data that support the findings of this study are openly available in BIMCV COVID-19+ at https://arxiv.org/pdf/2006.01174. The data that support the findings of this study are openly available in Radiological Society of North America at https://www.kaggle.com/c/rsna-pneumonia-detection-challenge. The data that support the findings of this study are openly available in Mendeley Data at https://data. mendeley.com/datasets/rscbjbr9sj/2.

Author Contribution Statement

Shubham: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing — original draft, Writing — review & editing, Visualization, Project administration. Brijesh Kumar Chaurasia: Conceptualization, Validation, Writing — review & editing, Visualization, Supervision, Project administration.