Multidimensional Analysis of Functional Near-Infrared Spectroscopy (fNIRS) Signal using Tucker Decomposition

Functional near-infrared spectroscopy (fNIRS) is a neuroimaging technique that could be used to research one's hemodynamic response to the world. A popular method to analyze the fNIRS signal is to use the grand averaging method, which collapses oxygenated hemoglobin (HBO) over a predefined time...

Full description

Saved in:
Bibliographic Details
Published inConference proceedings - IEEE International Conference on Systems, Man, and Cybernetics pp. 1317 - 1321
Main Authors Chan, Jasmine, Wilcox, Teresa, Hssayeni, Murtadha, Ghoraani, Behnaz
Format Conference Proceeding
LanguageEnglish
Published IEEE 09.10.2022
Subjects
Analysis of variance
Cybernetics
Functional near-infrared spectroscopy
Hemodynamics
infant
Neuroimaging
Signal analysis
tensor decomposition
Tensors
Tucker decomposition
Online AccessGet full text

Cover

More Information
Summary:Functional near-infrared spectroscopy (fNIRS) is a neuroimaging technique that could be used to research one's hemodynamic response to the world. A popular method to analyze the fNIRS signal is to use the grand averaging method, which collapses oxygenated hemoglobin (HBO) over a predefined time of interest (TOI) window. A drawback of the grand averaging method is that it collapses information about the fNIRS signal (e.g., temporal), thus one is not able to observe the temporal dynamics of the signal. Therefore, we propose to use multidimensional signal analysis (i.e., Tucker decomposition [TD] method) for fNIRS signal analysis, which can compress the signal and reveal the changes in the signal across time and space. We used TD on a three-way tensor with temporal, spatial, and subject modes constructed from an fNIRS dataset collected from infants that observed entities (i.e., human hand and mechanical claw) performing various action sequences (i.e., functional and nonfunctional events). Analysis of variance (ANOVA) was applied to the subject dimension to identify significant differences across conditions. We compared the performance of the grand averaging and TD method. Results from the TD method were able to replicate the results from the grand averaging method and identify additional patterns missed by the grand averaging method. Findings from this study demonstrate the TD method as an alternative fNIRS signal analysis method.
ISSN:2577-1655
DOI:10.1109/SMC53654.2022.9945539