Multiple correlation and multi-seed for robust inference of functional connectivity in FMRI

A novel statistical method for estimating brain networks from functional MRI data is presented. Instead of examining the correlations with each individual seed, we detect functional connectivity from fMRI data by simultaneously examining the multi-seed correlations via the multiple correlation coefficients. In addition, we propose to take into account the spatially structured noise in fMRI during the identification of the networks of functional interconnections by comparing the temporal multiple correlations against a model of the spatial multiple correlations in the noise. Evaluation for accuracy and robustness of the approach was performed using both simulated data and real fMRI data.