Department of computer, Sep.C., Islamic Azad University, Sepidan, Iran
10.48308/ijrtei.2026.241110.1099
Abstract
Purpose: The rapid advancements in convolutional neural networks (CNNs) have significantly improved medical image analysis. The COVID-19 pandemic has impacted millions worldwide, with containment hindered by inadequate testing resources and inefficiencies in diagnostic methods. This study proposes and explores a novel framework employing an ensemble of advanced deep transfer learning techniques for accurate and consistent COVID-19 detection from Computed Tomography (CT) scans, reducing reliance on manual assessment. Materials and Methods: The proposed framework integrates standardized data pre-processing with fine-tuned heterogeneous transfer learning models, including CNN- and Transformer-based architectures. An ensemble learning strategy is implemented at the feature level using Principal Component Analysis (PCA) to fuse deep representations extracted from the most effective models. The framework has been evaluated on two large publicly available CT datasets, COVID-CT and SARS-CoV-2, comprising over 1,600 COVID-19 and 1,450 non-COVID-19 images. Results: Experiments have demonstrated that fusing five architectures—ResNet-50 v2, EfficientNet-B5, ViT, VGG16, and DenseNet-201—achieves superior diagnostic performance compared to individual models and existing frameworks, as reflected by improved F1-scores. Conclusion: The results have confirmed that integrating transfer learning with feature-level ensemble learning within a unified framework significantly enhances the robustness and accuracy of COVID-19 detection from CT images. The proposed methodology provides a scalable and reproducible solution that can be extended to other medical image-based diagnostic tasks.
Ghasemi, S. G (2025). Ensemble of Transfer Learning Techniques for Detection of COVID-19 based on CT scans. International Journal of Research and Technology in Electrical Industry, 4(2), -. doi: 10.48308/ijrtei.2026.241110.1099
MLA
Ghasemi, S. G. "Ensemble of Transfer Learning Techniques for Detection of COVID-19 based on CT scans", International Journal of Research and Technology in Electrical Industry, 4, 2, 2025, -. doi: 10.48308/ijrtei.2026.241110.1099
HARVARD
Ghasemi, S. G (2025). 'Ensemble of Transfer Learning Techniques for Detection of COVID-19 based on CT scans', International Journal of Research and Technology in Electrical Industry, 4(2), pp. -. doi: 10.48308/ijrtei.2026.241110.1099
CHICAGO
S. G Ghasemi, "Ensemble of Transfer Learning Techniques for Detection of COVID-19 based on CT scans," International Journal of Research and Technology in Electrical Industry, 4 2 (2025): -, doi: 10.48308/ijrtei.2026.241110.1099
VANCOUVER
Ghasemi, S. G Ensemble of Transfer Learning Techniques for Detection of COVID-19 based on CT scans. International Journal of Research and Technology in Electrical Industry, 2025; 4(2): -. doi: 10.48308/ijrtei.2026.241110.1099
)