Privacy-preserving federated learning with optimized ensemble weighting and knowledge distillation for COVID-19 detection from non-IID medical imaging data

Medical imaging enables rapid and accurate diagnosis of COVID-19, with CT scans proving especially effective. However, data privacy concerns limit collaborative model development across hospitals. To address this issue, we introduce a novel federated learning framework. It is referred to as Independent Knowledge Distillation with post-Ensemble Federated Learning (IKDEFL). Differential Privacy (DP) is integrated into the framework to improve privacy guarantees. Three DP mechanisms are evaluated. These include Fixed Gaussian, Gaussian Adaptive, and Tree Adaptive. The evaluation has been conducted on heterogeneous and Non-Independent and Identically Distributed (Non-IID) datasets. These datasets reflect real-world hospital scenarios. Results show that IKDEFL significantly outperforms existing federated knowledge distillation methods. It achieves a generalization performance with accuracy up to 82.79% and an F1-score of 82.78% on Non-IID data. Among the DP methods, the Tree Adaptive mechanism has consistently provided the best trade-off between privacy and prediction quality. Peak accuracy reaches 76.62% under strict privacy constraints, where ε = 3.90 and δ = 10⁻³. This result is close to the performance of models without privacy protections. These findings demonstrate that adaptive DP techniques can be effectively applied in federated healthcare models. They support the development of privacy-preserving AI systems for clinical diagnostics.