Clustering Principal Components for Identification and Analysis of Prime Activity and Accompanying Activities in Scattered fMRI Data

The multi stimuli fMRI data tends to appear scattered because of involvement of multiple activity patterns. The detection of prime activity in the scattered data leads to decision making on prime stimulus. It also helps in finding the other accompanying activity patterns which aid the execution of prime activity. In this paper the detection of prime activity is performed through identification of principal components in the multivariate data. The principle components always align along the principal axis with the component with maximum variance nearest to the origin and the minimum variance component at the other extreme end. Clustering of the principle components yields groups of components with most similar variance values. The average linkage clustering is implemented for clustering the principle components. The cluster with maximum variance forms the principal component. This component represents the prime activity. The next maximum variance clusters represent the accompanying activities.