Reduction Of Noise Associated With Stimulus Correlated Motion In Event Related Overt Word Production fMRI Studies

function associated with the stimulus, y(t)= S(t)⊗H(t) + e(t) ,w here S(t) is the stimulus time series and H(t) is the IRF, e(t) is the gaussian noise and y(t) is the signal at a given voxel. Thus knowing the signal time series and the stimulus vector, one can estimate the IRF by the deconvolution of y(t) and S(t). The above analysis is very general and the IRF associated with the stimulus correlated motion can estimated along with the IRF for cortical activation. For an event related overt word production paradigm we and others [1] have found that the signal changes associated with the SCM are resolvable from the signal changes due to cortical activation. Hence the fMRI time series can be orthogonalized with respect to the SCM time series which is obtained by convolving the SCM related IRF with the stimulus vector. After reanalyzing the detrended fMRI time series we find that the functional activation map is much cleaner than that which is obtained without the orthogonalization.