Identification of brain activity from fMRI data: comparison of three fractal scaling analyses

Functional magnetic resonance imaging (fMRI) signal changes can be separated from background noise by various signal/image processing algorithms, including the well-known method, deconvolution. However, discriminating signal changes due to task-related brain activities from those due to task-related head motion, or due to other artifacts related in time to the task, has been little addressed. We examine whether three exploratory fractal scaling analyses that capture temporal correlation, the fluctuation analysis (FA), wavelet multiresolution analysis (WMA), and detrended fluctuation analysis (DFA), can be effective and reliable in this task. We find that DFA is indeed so. Brain activation maps derived by DFA are similar to maps derived by deconvolution. We also assess a signal model recently introduced by Birn et al. by applying DFA to simulation data generated from this model. We find that Birn's model fits our experimental data very well. Deconvolution explicitly uses information about task timing to extract the signals whereas DFA does not. Therefore, DFA is a promising practical method for fMRI data analysis