Hierarchical Clustering Of Independent Components Extracted From fMRI Data

Independent component analysis (ICA) of functional magnetic resonance imaging (fMRI) data can be employed as an exploratory method. The lack in the ICA model of strong a priori assumptions about the activation related signal or about the noise, leads to difficult interpretations of the results by the experimenters. Moreover, the statistical independence, hypothesized in the model, is only approximated by ICA algorithms. Residual dependencies among the extracted components can be investigated in order to reveal some informative structure in the data. In this work we propose a method based on hierarchical clustering algorithm in order to classify the results of ICA applied to fMRI dataset: the clustering algorithm uses a similarity measure based on mutual information between the extracted components. This method could be useful also to overcome the ambiguity related to the model order selection. The method was tested on simulated datasets. Preliminary results on real data are reported and discussed.