Data-driven analysis of hemodynamic response delay in event-related fMRI using wavelet transform

Functional magnetic resonance imaging (fMRI) provides high-resolution datasets that allow neuroscientists to sensitively detect activated region relating to a given task. Cerebral hemodynamic response (HR) of neural activity at the activated region is delayed and dispersed in time. This study proposes a new data-driven analysis for detecting the activated region and measuring the temporal delay of HRs. Our method can be divided into four steps. Firstly, a standard preprocessing (realigning and high pass filtering) is applied to the fMRI data. Secondly, we reconstruct the time course of signal in order to improve the time resolution. Thirdly, we derive the form and the delays of HRs using continuous wavelet transform. Finally, we give wavelet coefficient as the degree of activation. The proposed method has been applied to 7 right-handed subjects on right or left hand-gripping task. Experimental results showed that primary motor cortex (Ml) and the supplementary motor area (SMA) were detected as activated region, and means of temporal delays of HRs were right/left = 4.49/4.92 sec at SMA and 5.41/5.46 sec at Ml. Consequently, our method could classify the difference of HR delay between Ml and SMA.