Group representative brain connectivity model of episodic encoding using large fMRI dataset

Abstract

Episodic memory is autobiographical memory crucial to human cognitive function. Brain connectivity modeling is an essential tool for describing brain function of interest. While a large body of research has focused on analyzing functional brain connectivity, most studies has focused on a connectivity of an individual. Brain connectivity studies are usually based on functional Magnetic Resonance Imaging (fMRI) which often involved many subjects. Only a few researches have focused on group level connectivity. We employ a framework called Tigramite to estimate causal interaction and model episodic memory connectivity on Human Connectome Project (HCP) BOLD data. Regions of interest are identified by performing fMRI univariate group subtraction analysis on working memory task-fMRI data using subsequent memory test paradigm. From each resulting individual connectivity graphs, we construct a group representative graph by performing median aggregation on each corresponding edges and nodes of each individual graphs. However, there is no established method to measure a quality of the group representative graph. So, we propose the use of Levenshtein distance as a measurement to validate the group representative graph if it represents general characteristics of each individual graphs in the population.