[Paper Review] A Hybrid Federated Learning Based Ensemble Approach for Lung Disease Diagnosis Leveraging Fusion of SWIN Transformer and CNN

Asif H. Chowdhury, Md Fahim Ul Islam+4 more|arXiv (Cornell University)|Feb 19, 2026

COVID-19 diagnosis using AI0 citations

TL;DR

The paper proposes a hybrid federated learning ensemble that fuses CNN-based transfer learning models (VGG-19, Inception V3, DenseNet201) with SWIN Transformer to diagnose COVID-19, pneumonia, and normal chest X-rays, and evaluates in a federated setting.

ABSTRACT

The significant advancements in computational power cre- ate a vast opportunity for using Artificial Intelligence in different ap- plications of healthcare and medical science. A Hybrid FL-Enabled Ensemble Approach For Lung Disease Diagnosis Leveraging a Combination of SWIN Transformer and CNN is the combination of cutting-edge technology of AI and Federated Learning. Since, medi- cal specialists and hospitals will have shared data space, based on that data, with the help of Artificial Intelligence and integration of federated learning, we can introduce a secure and distributed system for medical data processing and create an efficient and reliable system. The proposed hybrid model enables the detection of COVID-19 and Pneumonia based on x-ray reports. We will use advanced and the latest available tech- nology offered by Tensorflow and Keras along with Microsoft-developed Vision Transformer, that can help to fight against the pandemic that the world has to fight together as a united. We focused on using the latest available CNN models (DenseNet201, Inception V3, VGG 19) and the Transformer model SWIN Transformer in order to prepare our hy- brid model that can provide a reliable solution as a helping hand for the physician in the medical field. In this research, we will discuss how the Federated learning-based Hybrid AI model can improve the accuracy of disease diagnosis and severity prediction of a patient using the real-time continual learning approach and how the integration of federated learn- ing can ensure hybrid model security and keep the authenticity of the information.

Motivation & Objective

• Develop a fusion model that combines transfer learning CNNs with a SWIN Transformer for lung disease detection.
• Leverage federated learning to enhance data privacy, reduce latency, and improve model reliability across hospitals.
• Evaluate performance on COVID-19 and pneumonia X-ray datasets and compare with individual models.

Proposed method

• Train CNN-based models (VGG-19, Inception V3, DenseNet201) and a SWIN Transformer separately on X-ray data.
• Ensemble the CNN models and fuse with SWIN Transformer to create a hybrid fusion model.
• Apply federated learning to distribute training across hospitals, aggregating local updates to update a global model.
• Measure performance using training/Testing times and accuracy; analyze ROC-AUC and confusion matrices.
• Compare fusion model against individual models and report results with sum and average fusion strategies.

Experimental results

Research questions

• RQ1 Does the hybrid fusion of CNN transfer learning models and SWIN Transformer improve diagnostic accuracy over individual models?
• RQ2 What are the gains in ROC-AUC and confusion-matrix metrics for the federated hybrid model compared to non-federated baselines?
• RQ3 How does federated learning affect training efficiency and resource usage in a multi-hospital setup?

Key findings

• Fusion model achieved 96.24% accuracy (sum fusion) and 94% (average fusion) on the tested dataset.
• Individual models achieved 94.4% (VGG-19), 94.5% (Inception V3), 94.1% (DenseNet201), and 82.5% (SWIN Transformer).
• SWIN Transformer alone underperformed relative to CNN models with 82.5% accuracy.
• Federated setup consumed substantial hardware resources (e.g., 35 GB RAM in a single run).
• Fusion model demonstrated higher ROC-AUC than individual models, with the fusion ensemble showing superior AUC in figures presented.
• Training times varied across models, with the sum fusion model showing the highest reported accuracy.