FED-EHR: A Privacy-Preserving Federated Learning Framework for Decentralized Healthcare Analytics
Makaleler > Tam Makale · 2025
Yazarlar: Rızwan Uz Zaman WANİ, Özgü CAN
ELECTRONICS (SWITZERLAND), SCI Expanded İndekslerine Giren Dergi, Uluslararası, Wos, 2025, cilt 14, ss 1-21
DOI: 10.3390/electronics14163261
Özet
<p>The Internet of Medical Things (IoMT) is revolutionizing healthcare by enabling continuous</p><p>monitoring and real-time data collection through interconnected medical devices such as</p><p>wearable sensors and smart health monitors. These devices generate sensitive physiological</p><p>data, including cardiac signals, glucose levels, and vital signs, that are integrated into electronic</p><p>health records (EHRs). Machine Learning (ML) and Deep Learning (DL) techniques</p><p>have shown significant potential for predictive diagnostics and decision support based</p><p>on such data. However, traditional centralized ML approaches raise significant privacy</p><p>concerns due to the transmission and aggregation of sensitive health information. Additionally,</p><p>compliance with data protection regulations, such as the Health Insurance Portability</p><p>and Accountability Act (HIPAA) and General Data Protection Regulation (GDPR), restricts</p><p>centralized data sharing and analytics. To address these challenges, this study introduces</p><p>FED-EHR, a privacy-preserving Federated Learning (FL) framework that enables collaborative</p><p>model training on distributed EHR datasets without transferring raw data from its</p><p>source. The framework is implemented using Logistic Regression (LR) and Multi-Layer Perceptron</p><p>(MLP) models and was evaluated using two publicly available clinical datasets: the</p><p>UCI Breast CancerWisconsin (Diagnostic) dataset and the Pima Indians Diabetes dataset.</p><p>The experimental results demonstrate that FED-EHR achieves a classification performance</p><p>comparable to centralized learning, with ROC-AUC scores of 0.83 for the Diabetes dataset</p><p>and 0.98 for the Breast Cancer dataset using MLP while preserving data privacy by ensuring</p><p>data locality. These findings highlight the practical feasibility and effectiveness of applying</p><p>the proposed FL approach in real-world IoMT scenarios, offering a secure, scalable, and</p><p>regulation-compliant solution for intelligent healthcare analytics.</p>