Abstract:
|
Resting state fMRI data generates dense functional connectivity networks that can be used to discriminate between neurologically differing groups of subjects. Functional connectivity is a measure of connectivity between brain regions sharing functional properties. This is measured by the correlation between regions of the brain, which results in a dense connectivity network. In attempting to find differences between groups, we treat each correlation as a variable in a regression model. This results in a difficult to analyze regression problem due to the fact that there are significantly more variables than the number of subjects. Therefore we investigate the use of sparse modeling algorithms such as LASSO, Sparse Discriminant Analysis, and others to reduce the number of variables in the model, while also focusing on the classification accuracy of distinguishing between the different subject groups.
|