RT Journal Article
SR Electronic
T1 Identification of Mechanisms of Functional Signaling Between Human Hippocampus Regions
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 099820
DO 10.1101/099820
A1 Ruben Sanchez-Romero
A1 Joseph D. Ramsey
A1 Jackson C. Liang
A1 Clark Glymour
YR 2017
UL http://biorxiv.org/content/early/2017/01/12/099820.abstract
AB Background Standard BOLD connectivity analyses depend on aggregating the signals of individual voxels within regions of interest (ROIs). In certain cases, this spatial aggregation implies a loss of valuable functional and anatomical information about subsets of voxels that drive the ROI level connectivity.New Method We use the FGES algorithm, a data-driven score-based graphical search method, to identify subsets of voxels that are chiefly responsible for exchanging signals between ROIs. We apply the method to high-resolution resting state functional magnetic resonance imaging (rs-fMRI) data from medial temporal lobe regions of interest of a single healthy individual measured repeated times over a year and a half.Results The FGES algorithm recovered subsets of voxels within larger medial temporal lobe ROIs of entorhinal cortex and hippocampus subfields that show spatially consistency across different scanning sessions, and are statistically significant under tests that validate the role of these subsets as main drivers of effective connectivity between hippocampal regions of interest.Comparison with Existing Methods In contrast to standard functional connectivity methods, the FGES algorithm is robust against false positive connections produced by transitive closures of adjacencies (correlation methods) and common effect conditioning (Markov random field methods).Conclusions The FGES algorithm allows for identification of communication subsets of voxels driving the connectivity between regions of interest, recovering valuable anatomical and functional information that is lost when ROIs are aggregated. The FGES algorithm is specially suited for voxelwise connectivity research, given its short running time and scalability to big data problems.