Tensor Based Temporal and Multilayer Community Detection for Studying Brain Dynamics During Resting State fMRI

• Published in: IEEE transactions on biomedical engineering Vol. 66; no. 3; pp. 695 - 709
• Main Authors: Al-sharoa, Esraa, Al-khassaweneh, Mahmood, Aviyente, Selin
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.03.2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Adult; Algorithms; Brain; Brain - diagnostic imaging; Brain - physiology; Brain architecture; Brain mapping; Cognitive ability; Communities; Community structure; Consistency; Dynamic functional connectivity networks; Dynamics; Evolution; Functional magnetic resonance imaging; Functional morphology; Hidden Markov models; Humans; Image Processing, Computer-Assisted - methods; Magnetic Resonance Imaging - methods; Male; Mathematical analysis; Matrix decomposition; Measurement; Multilayers; Network analysis; Networks; Neural networks; Organizations; Population studies; Principal component analysis; Rest - physiology; resting state fMRI (rs-fMRI); spectral clustering; Tensile stress; tensor decomposition; Tensors; Young Adult
• DOI: 10.1109/TBME.2018.2854676
• ISSN: 0018-9294; 1558-2531

Objective: In recent years, resting state fMRI has been widely utilized to understand the functional organization of the brain for healthy and disease populations. Recent studies show that functional connectivity during resting state is a dynamic process. Studying this temporal dynamics provides a better understanding of the human brain compared to static network analysis. Methods: In this paper, a new tensor based temporal and multi-layer community detection algorithm is introduced to identify and track the brain network community structure across time and subjects. The framework studies the temporal evolution of communities in fMRI connectivity networks constructed across different regions of interests. The proposed approach relies on determining the subspace that best describes the community structure using Tucker decomposition of the tensor. Results: The brain dynamics are summarized into a set of functional connectivity states that are repeated over time and subjects. The dynamic behavior of the brain is evaluated in terms of consistency of different subnetworks during resting state. The results illustrate that some of the networks, such as the default mode, cognitive control and bilateral limbic networks, have low consistency over time indicating their dynamic behavior. Conclusion: The results indicate that the functional connectivity of the brain is dynamic and the detected community structure experiences smooth temporal evolution. Significance: The work in this paper provides evidence for temporal brain dynamics during resting state through dynamic multi-layer community detection which enables us to better understand the behavior of different subnetworks.