Precision white-matter connectomes to study individuality and variability in human populations

FRANCO PESTILLI; CESAR F. CAIAFA; BRENT MCPHERSON

Ginebra

Conferencia; 22nd Annual Meeting of the Organization for Human Brain Mapping; 2016

The fundamental inner workings of the brain cannot be simply understood at the level of single neurons or brain areas. To capture human behavior, we need to understand the brain's circuit behavior. One of the next frontiers in brain science will be to develop models and theories that allow investigators to understand how networks support brain function in individual brains. Diffusion-weighted magnetic resonance imaging (dMRI) and fiber trackingallow mapping the network structure of brain connections (white matter tracts) in living brains; the connectome. By measuring white matter anatomy and connections in living brains these technologies allow charting the relation between human behavior, development, and aging with brain function and structure. Yet efforts will be necessary to improve accuracy in connectome estimation.Recently, we developed a method (LiFE, Linear Fascicle Evaluation; Pestilli et al., 2014) for computing connectome accuracy. LiFE computes how well a connectome predicts the measured diffusion signal. It can be used to identify the most accurate connectomes in individual brains. Here we exploit recent multi-way factorization methods (Cichocki, et al, 2015) to build a new version of the LiFE model. We show that the factorized LiFE model hasaccuracy indistinguishable from the original LiFE model without requiring large computer memory. In addition, we combine the factorized LiFE with innovative Ensemble Tracking methods (ET; Takemura, et al., 2016) to generate a new approach to individualized in-vivo connectomics. We build ET connectomes by combining multiple tractography methods (i.e.., deterministic and probabilistic tracking based on both tensor and constrained spherical deconvolution models; Tournier et al., 2012 ). We show that the new approach generates ETconnectomes that are highly reliable within an individual (i.e., they are reproducible across test-retest), but maintain anatomical features that allow fine discrimination between.