Regionally specific sensitivity differences in fMRI and PET: where do they come from?

In this paper we report three neuroimaging studies of language that investigate potential sources of inconsistency in measured hemodynamic responses: (1) between sessions for fMRI, including differences in hormonal status, (2) between sessions for PET, and (3) between scanning modalities (PET and fMRI). Differences in evoked responses between sessions of the same modality were small. In particular we did not find any effect of hormone levels when testing during the first and third weeks of the menstrual cycle (although we cannot exclude the possibility that activation in the temporoparietal regions is sensitive to hormonal status). Comparing the two modalities showed that prefrontal regions were more activated in fMRI than in PET. This may relate to task switching between blocks in fMRI that is not induced by PET paradigms or increased error variance in these regions for PET. In contrast, temporal activations were found in PET more than in fMRI. We attribute the lack of temporal activations, in fMRI, to a combination of factors, including susceptibility artifacts, anticipatory activity during the control condition, discontinuous sampling of peristimulus time, and differences in the source, acquisition, and analysis of the measured signals. It is concluded that although there is sufficient reproducibility of results for these paradigms within each modality, the regionally specific differences in sensitivity found between modalities warrant further investigation. These regionally specific differences are important for a properly qualified interpretation of activation profiles in fMRI. Copyright 2000 Academic Press.