Precision mapping of structural connectomes in individual brains

OLAF SPORNS; BRENT MCPHERSON; FRANCO PESTILLI; CESAR F. CAIAFA

San Diego

Conferencia; Society for Neuroscience SfN 2016; 2016

SfN

Diffusion weighted Magnetic Resonance Imaging (dMRI) and subsequent tractography-based modeling allows characterizing and quantifying connectivity and tissue properties of white-matter in living brains. By integrating this fiber-tracking information with a gray matter parcellation, it is possible to model a human connectome as a structural connectivity matrix. Such network-based representation allows for testing a wide number of integration, segregation, and overall communicability features, such as clustering coefficient and small worldness, which have been compared to differences in behavior, cognition, development and aging processes across groups. To date, connectome research has been concerned primarily with identifying principles of brain function and structure by averaging connectomes from multiple subjects. Recently, renewed interest has emerged in developing precision methods to reliably map individual connectomes. Mapping connectomes reliably in single brains would allow measuring individuality and variability across humans as well as tracking longitudinally connectome variations within individuals in an accurate way. In this work, we introduce methods to map statistically-validated structural connectomes. We report results on the reliability of connectome estimates and network measures. We used two datasets dMRI (Van Essen et al. 2013; Pestilli et al. 2014) and multiple fiber tracking methods (Tournier et al. 2012; Descoteaux et al. 2009; Basser et al. 2000). We built ten connectomes in each individual brain and applied the Linear Fascicle Evaluation (LiFE; (Pestilli et al. 2014) method to validate each connectome. Whole brain connectomes were generated avoiding fiber-counts measures but using virtual lesions (Pestilli et al. 2014) combined with FreeSurfer (Fischl 2012). To estimate the precision of the LiFE-validated connectomes, we measured the reliability of standard brain network-properties such as clustering coefficient and small worldness across repeated instances of connectomes build in the same individual using each tracking method. We show the degree to which LiFE-based connectomes achieve reliable levels of replicability within repeats of individual subjects while maintaining good levels of between-subject discriminability as observed by within- and between-group variability. Our results show that structural connectomes processed with LiFE offer more reliable estimates of connections between brain regions. The method contributes to precision connectome mapping methods with the potential to impact detection during prodromal stages of disease or for individuals at genetic risk of disease.