Clustering functional magnetic resonance imaging (fMRI) time series has eme
rged in recent years as a possible alternative to parametric modeling appro
aches. Most of the work so far has been concerned with clustering raw time
series. In this contribution we investigate the applicability of a clusteri
ng method applied to features extracted from the data. This approach is ext
remely versatile and encompasses previously published results [Goutte et al
., 1999] as special cases. A typical application is in data reduction: as t
he increase in temporal resolution of fMRI experiments routinely yields fMR
I sequences containing several hundreds of images, it is sometimes necessar
y to invoke feature extraction to reduce the dimensionality of the data spa
ce. A second interesting application is in the meta-analysis of fMRI experi
ment, where features are obtained from a possibly large number of single-vo
xel analyses. Ln particular this allows the checking of the differences and
agreements between different methods of analysis. Both approaches are illu
strated on a fMRI data set involving visual stimulation, and we show that t
he feature space clustering approach yields nontrivial results and, in part
icular, shows interesting differences between individual voxel analysis per
formed with traditional methods. (C) 2001 Wiley-Liss, Inc.