Anatomic and functional variability: the effects of filter size in group fMRI data analysis

In the analysis of group fMRI scans, an optimal spatial filter should be large enough to accurately blend functionally homologous anatomic regions, yet small enough not to blur the functionally distinct regions. Hanning filters varying from 0.0 to 18.0 mm were evaluated in a group analysis of six healthy controls performing a simple finger-tapping paradigm. Test-retest reliability and Talairach-based measurements of the sensorimotor region were used to explore the optimal filter size. Two distinct regions of functional activation were noted in the sensorimotor cortex in group images (n = 6) at both time 1 and time 2. These regions merge once the filter size exceeds approximately 6.0 mm. The original hypothesis that these represented a motor and sensory activation was rejected on the basis of structural and functional variability. A discussion of the inherent difficulties in choosing an appropriate filter size is presented.