Variability of Functional Connectomes Through Community Structure

• Published in: 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW) pp. 1 - 5
• Main Authors: Osterkamp, Brooke, Ortiz-Bouza, Meiby, Aviyente, Selin
• Format: Conference Proceeding
• Language: English
• Published: IEEE 04.06.2023
• Subjects: Brain modeling; Community Detection; Electroencephalogram; Feature extraction; Fingerprint recognition; Functional Connectome; Image edge detection; Individual Variability; Multiplexing; Predictive models; Signal processing
• DOI: 10.1109/ICASSPW59220.2023.10193610

Traditional neuroimaging research typically focuses on identifying group level characteristics of brain connectivity across subjects. However, recent research has shown that even among the neurologically healthy, brain function shows high individual variability. This realization has lead to the concept of functional connectome (FC) fingerprinting, where the functional connectivity profiles act as a fingerprint that can be used to identify the individual and establish a relationship between brain function and behavior. A core component of fingerprinting is feature selection where a subset of edges from FC are selected. In this paper, we propose an alternative feature selection approach where functional subnetworks instead of individual edges are identified to describe group and individual level network activity. The proposed approach models the FCs from multiple subjects as a multiplex network and employs a community detection method to extract functional subnetworks that are common and unique across subjects. We then introduce metrics to evaluate the common communities and quantify their consistency and variability across subjects. The proposed framework is evaluated on functional connectomes constructed from electroencephalogram (EEG) data during a cognitive control study.