A Quantum-Enhanced Federated Deep Learning Framework for Privacy-Preserving Early Diagnosis of Multi-Organ Diseases Using Multi-Modal Electronic Health Records

Abstract

The centralization of deep learning is one of the key concerns when using Electronic Health Records (EHRs) to diagnose multi-organ diseases early to enhance clinical outcomes because of severe privacy and data-sharing issues. The proposed research offers a Quantum-Enhanced Federated Deep Learning (QEFDL) framework, which is privacy-preserving early diagnosis with multi-modal EHR, such as clinical records, laboratory results, medical images, and demographic information. The framework merges quantum-assisted multimodal feature fusion with multimodal feature fusion and secure federated learning that relies on differential privacy and secure aggregation. The proposed model was tested with a distributed dataset of 120,000 patient records in various healthcare facilities. Experimental results achieved an accuracy of 98.72%, precision of 98.35%, recall of 98.18%, F1-score of 98.26%, and an AUC of 0.992. Moreover, the framework minimized communication overheads (by 27.8%), convergence rounds (by 31.4%), and privacy leakage (by 91.3%), employing it to illustrate its applicability in ensuring secure, scalable and accurate AI-based clinical decision support.