Variability of Non-parametric HRF in Interconnectedness and Its Association in Deriving Resting State Network

Document Type

Conference Proceeding

Publication Date

1-1-2023

Abstract

Blood Oxygen Level-Dependent (BOLD) time course in functional magnetic resonance imaging (fMRI) is modeled as the response of the hemodynamic response function (HRF) excited by an activity-inducing signal. Variability of the HRF across the brain influences functional connectivity (FC) estimates and some approaches have been attempted to separate the HRF and activity-inducing signal from the observed BOLD signal as a blind separation problem. In this work, an approach based on homomorphic filtering is proposed to estimate a non-parametric representation of HRF in resting state fMRI. Voxel-wise and region-wise variations of correlation of the estimated HRF (both the parametric and non-parametric representation) are analyzed in different functional networks. Principal component analysis of the correlation matrix using the estimated HRF is used to analyze the interconnectedness. HRF shows higher variability for the non-parametric representation over the parametric representation. Further, the contribution of the estimated HRF is then studied in producing resting-state networks using the dictionary learning framework.

Identifier

85172416379 (Scopus)

ISBN

[9783031430749]

Publication Title

Lecture Notes in Computer Science Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics

External Full Text Location

https://doi.org/10.1007/978-3-031-43075-6_21

e-ISSN

16113349

ISSN

03029743

First Page

239

Last Page

248

Volume

13974 LNAI

Grant

2021B1515020019

Fund Ref

National Science Foundation

This document is currently not available here.

Share

COinS