Structured Robust Tensor Decomposition for Multilayer Community Detection: Application to Functional Connectivity Networks

Functional organization of the brain can be modeled as a network of interconnected regions. Study of brain networks has offered new insights to human behavior and neurodegenerative diseases. While traditional approach to functional connectivity focuses on average connectivity within a given time win...

Full description

Saved in:
Bibliographic Details
Published in2024 32nd European Signal Processing Conference (EUSIPCO) pp. 1342 - 1346
Main Authors Al-Sharoa, Esraa, Alwardat, Mohammad, Aviyente, Selin
Format Conference Proceeding
LanguageEnglish
Published European Association for Signal Processing - EURASIP 26.08.2024
Subjects
Diseases
EEG
Electroencephalography
Europe
functional connectivity
Functional magnetic resonance imaging
low-rank
Monitoring
Neuroimaging
Nonhomogeneous media
Organizations
Robust tensor decomposition
Signal processing
spectral clustering
Tensors
Online AccessGet full text

Cover

More Information
Summary:Functional organization of the brain can be modeled as a network of interconnected regions. Study of brain networks has offered new insights to human behavior and neurodegenerative diseases. While traditional approach to functional connectivity focuses on average connectivity within a given time window, recent research shows that the temporal dynamics of functional connectivity (FC) change over multiple time scales during task performance and rest. In this paper, we present a structured robust low-rank tensor decomposition to detect the community structure of functional connectivity networks across multiple subjects and time. The proposed approach combines robust tensor decomposition with a spectral clustering regularization term to extract the group level community structure at each time point. Moreover, the contribution of each subject to this community structure is quantified by the learned weighting coefficients. The proposed framework is applied to functional connectivity networks constructed from task-based electroen-cephalogram (EEG) data.
ISSN:2076-1465
DOI:10.23919/EUSIPCO63174.2024.10715157