Related Experiment Videos
Partial least squares analysis of neuroimaging data: applications and advances
Anthony Randal McIntosh1, Nancy J Lobaugh
1Rotman Research Institute of Baycrest Centre, University of Toronto, Toronto, Ontario, Canada M6A 2E1. mcintosh@psych.utoronto.ca
Neuroimage
|October 27, 2004
Summary
Partial least squares (PLS) analysis is a powerful tool for understanding brain activity across neuroimaging methods like fMRI and ERP. This review details its application to spatiotemporal patterns, offering insights into brain function.
Area of Science:
- Neuroscience
- Cognitive Science
- Data Analysis
Background:
- Partial least squares (PLS) is a statistical method used to analyze distributed signals from neuroimaging techniques.
- Previous applications focused on positron emission tomography (PET) for tasks, behavior, and functional connections.
Purpose of the Study:
- To review the extension of PLS for analyzing spatiotemporal patterns in functional magnetic resonance imaging (fMRI), event-related potentials (ERP), and magnetoencephalography (MEG).
- To provide a mathematical description and discuss statistical assessment methods for PLS in neuroimaging.
Main Methods:
- Mathematical description of PLS.
- Statistical assessment using permutation testing and bootstrap resampling.
- Illustration with simulated and empirical ERP and fMRI data.
Main Results:
- PLS can characterize spatiotemporal patterns in fMRI, ERP, and MEG data.
- Permutation testing assesses statistical strength of activity patterns.
- Bootstrap resampling evaluates the reliability of regional contributions.
Conclusions:
- Spatiotemporal PLS is a valuable tool for neuroimaging analysis, particularly for fMRI and ERP data.
- Statistical assessment via permutation testing and bootstrapping enhances result interpretation.
- PLS should be considered within a broader, pluralistic approach to neuroimaging data analysis.