Characterizing nonlinear relationships in functional imaging data using eigenspace maximal information canonical correlation analysis (emiCCA)

A novel unsupervised and data-driven method termed eigenspace maximal information canonical correlation analysis (emiCCA) and a framework of fMRI data analysis using emiCCA are proposed. The crucial point of our work was to utilize the eigenvectors and eigenvalues from the eigenspaces of the maximal information coefficient (MIC) matrix as a new measure for assessing the relationships between two data sets.294