Susceptibility-induced loss of signal: comparing PET and fMRI on a semantic task

Functional magnetic resonance imaging (fMRI) has become a popular tool for investigations into the neural correlates of cognitive activity. One limitation of fMRI, however, is that it has difficulty imaging regions near tissue interfaces due to distortions from macroscopic susceptibility effects which become more severe at higher magnetic field strengths. This difficulty can be particularly problematic for language tasks that engage regions of the temporal lobes near the air-filled sinuses. This paper investigates susceptibility-induced signal loss in the temporal lobes and proposes that by defining a priori regions of interest and using the small-volume statistical correction of K. J. Worsley, S. Marrett, P. Neelin, A. C. Vandal, K. J. Friston, and A. C. Evans (1996, Hum. Brain Mapp. 4: 58-83), activations in these areas can sometimes be detected by increasing the statistical power of the analysis. We conducted two experiments, one with PET and the other with fMRI, using almost identical semantic categorization paradigms and comparable methods of analysis. There were areas of overlap as well as differences between the PET and fMRI results. One anticipated difference was a lack of activation in two regions in the temporal lobe on initial analyses in the fMRI data set. With a specific region of interest, however, activation in one of the regions was detected. These experiments demonstrate three points: first, even for almost identical cognitive tasks such as those in this study, PET and fMRI may not produce identical results; second, differences between the two methods due to macroscopic susceptibility artifacts in fMRI can be overcome with appropriate statistical corrections, but only partially; and third, new data acquisition paradigms are necessary to fully deal with susceptibility-induced signal loss if the sensitivity of the fMRI experiment to temporal lobe activations is to be enhanced.