File(s) under permanent embargo
A least angle regression method for fMRI activation detection in phase-encoded experimental designs
journal contribution
posted on 2023-06-09, 02:33 authored by Xingfeng Li, Damien Coyle, Liam Maguire, Thomas M McGinnity, David R Watson, Habib BenaliThis paper presents a new regression method for functional magnetic resonance imaging (fMRI) activation detection. Unlike general linear models (GLM), this method is based on selecting models for activation detection adaptively which overcomes the limitation of requiring a predefined design matrix in GLM. This limitation is because GLM designs assume that the response of the neuron populations will be the same for the same stimuli, which is often not the case. In this work, the fMRI hemodynamic response model is selected from a series of models constructed online by the least angle regression (LARS) method. The slow drift terms in the design matrix for the activation detection are determined adaptively according to the fMRI response in order to achieve the best fit for each fMRI response. The LARS method is then applied along with the Moore– Penrose pseudoinverse (PINV) and fast orthogonal search (FOS) algorithm for implementation of the selected model to include the drift effects in the design matrix. Comparisons with GLM were made using 11 normal subjects to test method superiority. This paper found that GLM with fixed design matrix was inferior compared to the described LARS method for fMRI activation detection in a phased-encoded experimental design. In addition, the proposed method has the advantage of increasing the degrees of freedom in the regression analysis. We conclude that the method described provides a new and novel approach to the detection of fMRI activation which is better than GLM based analyses.
History
Publication status
- Published
File Version
- Published version
Journal
NeuroImageISSN
1053-8119Publisher
ElsevierExternal DOI
Issue
4Volume
52Page range
1390-1400Department affiliated with
- Division of Medical Education Publications
Full text available
- No
Peer reviewed?
- Yes
Legacy Posted Date
2016-08-16First Compliant Deposit (FCD) Date
2016-08-16Usage metrics
Categories
No categories selectedKeywords
Licence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC