BCFTL: Blockchain-Enabled Multimodal Federated Transfer Learning for Decentralized Alzheimer's Diagnosis
Myrzashova, Raushan Alsamhi, Saeed Hamood Shvetsov, Alexey V. Hawbani, Ammar Guizani, Mohsen Wei, Xi
Myrzashova, Raushan
Alsamhi, Saeed Hamood
Shvetsov, Alexey V.
Hawbani, Ammar
Guizani, Mohsen
Wei, Xi
Supervisor
Department
Machine Learning
Embargo End Date
Type
Journal article
Date
2025
License
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
Efficient monitoring of carbon capture and storage (CCS) systems heavily relies on sensor data. However, sensors are susceptible to potential multiple faults, leading to performance degradation and posing a risk of catastrophic failures. Timely detection of sensor faults in CCS systems is crucial for safe and efficient carbon dioxide (CO2) pipeline operation. This paper addresses the challenge of diagnosing multiple sensor faults in CO2 pipelines by introducing a novel approach based on partial-distributed particle filter (PDPF). The novel distributed-filtering framework aims to reduce the computational complexity while identifying multiple faults in highly nonlinear systems. The proposed PDPF architecture comprises a collection of linear local filters and a nonlinear main filter. More specifically, the algorithm segregates nonlinear computations from local filters and assigns them to the main filter. The main filter handles the time updates involving all nonlinear computations associated with the nonlinear system, while the parallel linear local filters, each equipped with a distinct subset of sensor measurements, perform the measurement updates and combine their estimates via information fusion. As for fault detection and isolation, each local filter utilizes a novel kernel density estimation (KDE)-based approach that analyzes the consistency between model predictions and observed behavior, enabling the identification of sensor faults. Compared to existing methods, this approach reduces the computational requirements and is well-suited for highly nonlinear systems experiencing multiple sensor faults. Additionally, performance assessment via numerical simulations confirms its effectiveness and superiority in comparison to stateof-the-art alternative methods. © 2014 IEEE.
Citation
R. Myrzashova, S. H. Alsamhi, A. V. Shvetsov, A. Hawbani, M. Guizani and X. Wei, "BCFTL: Blockchain-Enabled Multimodal Federated Transfer Learning for Decentralized Alzheimer’s Diagnosis," in IEEE Internet of Things Journal, doi: 10.1109/JIOT.2025.3569652
Source
IEEE Internet of Things Journal
Conference
Keywords
Alzheimer, Blockchain, Clinical Records, Decentralized Data Sharing, Federated Learning, MRI, Multimodal, Remote Healthcare, Transfer Learning
Subjects
Source
Publisher
IEEE