Dr. Christian Habeck
"Multivariate approaches to neuroimaging analysis 101"
Friday, November 2, 2007
As the clinical and cognitive neurosciences mature, multivariate analysis techniques for neuroimaging data have received increasing attention since they possess some attractive features that cannot be easily realized by the more commonly used univariate, voxel-wise, techniques: (1) multivariate approaches evaluate correlation/covariance of activation across brain regions, rather than proceeding on a voxel-wise basis. Thus, their results can be more easily interpreted as a signature of neural networks. (2) Multivariate techniques also lend themselves better to prospective application of results obtained from the analysis of one dataset to entirely new datasets. Multivariate techniques are thus well placed to provide information about mean differences and correlations with behavior, similarly to univariate approaches, but with potentially greater statistical power and better reproducibility checks. Despite these attractive features, the barrier of entry to the use of multivariate approaches has been high, preventing more widespread application in the community. We have therefore proposed a series of studies comparing multivariate approaches amongst each other and with traditional univariate approaches in didactic reports and comprehensive review papers, using simulated as well as real-world data sets.
In my presentation, I will give a simple mathematical overview of univariate and multivariate approaches. Then I‚ll present two examples of multivariate analysis applied to: (1) an fMRI data set from a study using a delayed-response task, and (2) a clinical data set from a study comparing healthy elderly participants with early Alzheimer‚s disease patients, using resting scans of regional Cerebral Blood Flow. The presentation will close with remaining challenges and directions for the future.