LINEAR AND NONLINEAR MODELING OF THE HEMODYNAMIC SYSTEM IN fMRI

Neuronal activity in the brain triggers variations in the blood oxygenation level, which can be measured through functional mri (fmri). thus the hemodynamic acts as a transitional system between brain activity and fmri signal. conventional models of the hemodynamic response function (hrf) are gaussian and gamma functions. they have parameters that are usually selected a priori, and may be different in various subjects. the novel work of this paper is the develompemnt of a new method that characterizes the hrf (over a wide range of variations) using three elementary functions, derived from singular value decomposition. we use these three elementary signals along with the stimulation pattern to develop both linear and nonlinear models of hemodynamic system. the new strategy can also be used for identification of hemodynamic system and in various activation detection methods. we apply it to activation detection in both event-related and block design experimental fmri data and also in both linear and nonlinear modeling of hrf. the activated regions detected via proposed models are consistent with previous studies, indicating the ability of the proposed approach for detecting activated brain regions without a priori assumptions about the shape parameters of the hrf. the capability of the proposed elementary functions for hrf identification is demonstrated through estimating the hrf in some of activated regions.