Neurocomputing 44–46 (2002) 1065–1070 www.elsevier.com/locate/neucom Functional connectivity by cross-correlation clustering Silke Dodel ∗ , J. Michael Herrmann, Theo Geisel Max-Planck-Institut f ur Str omungsforschung, Bunsenstrasse 10, 37073 G ottingen, Germany Abstract In addition to information on localization of brain functions, data from fMRI experiments con- tain also cues about the functional connectivity among modular units. We propose a data-driven deterministic clustering algorithm based on temporal cross-correlations and elements of graph theory to detect functionally connected regions. The cluster concept can be changed in a con- trolled manner to reveal the functional connectivity structure in detail. The algorithm is applied to data from a motor task and shows to successfully determine clusters related to the stimulus. Furthermore, the method can be extended to include the analysis of temporal relations between dierent brain regions. c  2002 Published by Elsevier Science B.V. Keywords: Functional connectivity; Cluster; Cross-correlation; Graph; Connectivity component; Clique; fMRI 1. Introduction Functional connectivity is an issue of increasing interest in functional magnetic reso- nance imaging (fMRI). Here, much more than for the identication of stimulus-related activity, data driven aspects of analysis methods are relevant. Several approaches to determine functional connectivity are known such as fuzzy clustering [3], dynamical cluster analysis [2] and combinations of various data driven methods [1]. We propose a method we call cross-correlation clustering which has the advantage of being determin- istic in that there is no need to make any prior assumptions on seeds or the number of clusters. The metric for the distance between two voxels of an fMR image is based on the cross-correlation of their signal time courses. The underlying assumption is that two or more voxels could be considered as functionally connected if the cross-correlation * Corresponding author. Tel.: +49-551-5176420; fax: +49-551-5176439. E-mail addresses: silke@chaos.gwdg.de (S. Dodel), michael@chaos.gwdg.de (J. M. Herrmann), geisel@chaos.gwdg.de (T. Geisel). 0925-2312/02/$-see front matter c  2002 Published by Elsevier Science B.V. PII:S0925-2312(02)00416-2