Architecture of Decentralized Expert System for Early Alzheimer's Prediction Enhanced by Data Anomaly Detection

Stefan Behfar, Qumars Behfar, marzie hosseinpour

Keywords: Early-stage AD prediction, Anomaly detection, Decentralized expert system, Alzheimer's disease, Blockchain

TL;DR: Decentralized Expert System for Early Alzheimer's Prediction

Abstract: Alzheimer’s Disease poses a significant global health challenge, necessitating early and precise detection to enhance patient outcomes. Traditional diagnostic methodologies often result in delayed and imprecise predictions, particularly in the disease’s early stages. Centralized data repositories struggle to manage the immense volumes of MRI data, alongside persistent privacy concerns that impede collaborative efforts. This paper presents an innovative approach that leverages the synergy of blockchain technology (due to crowdsourcing patients' longitudinal test data via Web3 application) and Federated Learning to address these challenges. Thus, our proposed decentralized expert system architecture presents a pioneering step towards revolutionizing disease diagnostics. Furthermore, the system integrates robust anomaly detection for patient-submitted data. It emphasizes AI-driven MRI analysis and incorporates a sophisticated data anomaly detection architecture. These mechanisms scrutinize patient-contributed data for various issues, including data quality problems. We acknowledge that performing an exhaustive check of the correctness and quality of MRI images and biological information directly on-chain is not practical due to the computational complexity and cost constraints of blockchain platforms. Instead, such checks are typically performed off-chain, and the blockchain is used to record the results securely. This comprehensive approach empowers to provide more precise early-stage Alzheimer’s Disease prediction with more volume of data. Our system is designed to safeguard both data integrity and patient privacy, facilitating collaborative efforts.

Primary Area: Machine learning for healthcare

Submission Number: 12088