Predicting stimulus-rate sensitivity of human somatosensory fMRI signals with MEG

With increasing stimulus rate (SR), cortical EEG and MEG responses typically decrease in amplitude whereas BOLD fMRI signals increase. To address this discrepancy, we predicted BOLD responses with squared MEG waveforms using a recently proposed energy-density model. Tactile stimuli were delivered to finger tips at SRs of 1, 4, or 10 Hz in successive 25-s blocks, and brain signals were detected from area 3b of the primary somatosensory cortex of nine healthy adults using a 306-channel whole-scalp neuromagnetometer and a 3-T fMRI magnet. The main MEG deflections decreased in amplitude as a function of SR, whereas the BOLD signals increased from 1- to 4-Hz SR, with no further change at 10 Hz. MEG energy densities, obtained over the whole stimulus train and convolved with different hemodynamic response functions, predicted both the shape and amplitude of the BOLD signals well, and incorporation of nonlinear terms into the model did not offer any further advantage. Thus, squared MEG waveforms obtained over the entire stimulus train provided an appropriate estimate of area 3b neuronal activity associated with the BOLD signal.