Abstract
Conventional analysis of functional magnetic resonance imaging (fmri) time series is based on univariate statistical analysis. In this approach, a spatially invariant model of the expected blood oxygenation level-dependent (bold) response is fitted independently at each voxel’s time course, and the differences between estimated activation levels during two or more experimental conditions are tested. Together with methods for mitigating the problem of performing a large number of tests, this massively univariate analysis produces statistical maps of response differences, highlighting brain locations that are “selective” or “specialized” for a certain stimulus dimension, that is, voxels or regions of interest (roi) that respond more vigorously to a sensory, motor, or cognitive stimulus compared to one or more appropriate control conditions. This approach is not appropriate when the relevant question is what is the “information” content of a certain network of brain regions rather than which is its “activation” level.keywordsfeature selectionindependent component analysistraining datasetunsupervised learningstimulus processingthese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.
| Original language | English |
|---|---|
| Title of host publication | fMRI: From Nuclear Spins to Brain Functions |
| Editors | Kamil Ugurbil, Kamil Uludag, Lawrence Berliner |
| Publisher | Springer |
| Pages | 661-698 |
| ISBN (Electronic) | 978-1-4899-7591-1 |
| ISBN (Print) | 978-1-4899-7590-4, 978-1-4899-7872-1 |
| DOIs | |
| Publication status | Published - 2015 |