Adapted estimate of neural activity based on blood-oxygen-level dependent signal by a model-free spatio-temporal clustering analysis

In this study, we detected brain activity by comparing the overall temporal response of the blood oxygen level referring to hemodynamic response with a modeled hemodynamic response (MHR). However, in a conventional analysis by statistical parametric mapping (SPM) method, the MHR is assumed to be a fixed-response function, which may bias the conclusions about brain activation, such as the shapes of the response curve or the different response delays to stimuli. Therefore, to improve detection efficacy, we applied a spatio-temporal clustering analysis (sTCA) to determine the MHR, which is calculated from the prospective voxels with no a priori information about the experiment design. With the sTCA method, these prospective voxels are detected by the feature with the largest temporal clustering within which these voxels react simultaneously, irrespective of where the variant hemodynamic response occurs. This estimated MHR (eMHR) is then applied to search for brain activation. Preliminary results show that the eMHR signal response closely resembles the real signal response of the target area. Moreover, the activation detection using eMHR method is more sensitive for the human visual and motor tasks than that with the canonical hemodynamic response embedded in the SPM analysis as the default MHR (dMHR). The more precise location of brain activation made possible by the improved sensitivity should provide helpful information about the stimulation of neuron activity.