Federated Edge Intelligence for Privacy-Preserving Pre-Eclampsia Prediction in IoT-Based Maternal Care

Abstract

Pre-eclampsia, a leading cause of maternal-fetal mortality, demands urgent innovation in prenatal monitoring. This paper presents an IoT-based wearable system integrating maternal (PPG, temperature) and fetal (kick) sensors with edge AI for real-time pre-eclampsia detection. Leveraging federated learning and Hyperledger blockchain, our solution ensures privacy-preserving analytics (82% lower data exposure) while achieving 94.2% accuracy at 200ms latency, 23 times faster than cloud alternatives. The system employs an ESP32-powered wearable with LoRaWAN connectivity, optimized via quantized Light GBM for 0.8J/inference energy efficiency. Clinical trials with 350 subjects in rural settings validated its scalability, with blockchain-secured aggregation handling 350 transactions/second. Key innovations include multi-modal sensor fusion for holistic risk assessment, on device TinyML eliminating cloud dependence, and tamper proof audit trails for HIPAA compliance. This work bridges critical gaps in latency, privacy, and fetal-maternal monitoring, offering a deployable solution for low-resource regions. Results demonstrate more than 90% alignment with clinician diagnoses, positioning our system as a transformative tool for early pre-eclampsia intervention.

Ähnliche Arbeiten