jovo
commented
7 years ago This dataset consists of $n=42$ subjects, each with $197$ time-steps of resting-state functional MRI activity, as well as the subject's five-factor personality trait as quantified by the NEO Personality Inventory-Revised \cite{Costa1992}. For the brain activity modality, we derived the following comparison function. For each scan, (i) run the Configurable Pipeline for the Analysis of Connectomes (C-PAC) pipeline \cite{CPAC2015} to process the raw brain images yielding a parcellation into $197$ regions of interest, (ii) run a spectral analysis on each region, (iii) bandpass and normalize it to sum to one, and (iv) calculate the Kullback-Leibler divergence across regions to obtain a similarity matrix across comparing all regions. Then, use the normalized Hellinger distance to compute distances between each subject.
@jovo