Direct measure of local region functional connectivity by multivariate correlation technique.

In order to identify the local areas whose activity are most similar with region of interest (ROI), we usually compute the correlation of fMRI data for the brain functional connectivity. The fMRI data is usually noisy, extraction of functional connectivity with the voxel by voxel based method such as Pearson correlation analysis is not robust. Many people smooth the fMRI data before compute the correlation coefficient, which only makes the effect worse, because some useful original information is lost during the smoothing. Here, we analyzed this issue in details and improved the data processing flow to make the result better. Furthermore, a new criterion RV correlation coefficient was introduced in this article to measure the correlation between two local brain regions; this multivariate correlation technique applied the spatiotemporal information within the local regions to measure the similarity of the activity in different brain regions. We compared four different strategies mentioned above to detect the functional connectivity on the simulated and real fMRI data, and the results demonstrated that the RV-coefficient method obtained the best performance.