Connectome mapping algorithms with application to community services for big data neuroscience

Neuroscience is on a search for methods for precision study of brain networks using neuroimaging data on large-scale human populations. We propose utilizing innovative technologies in human connectomics, which can delineate the architecture of brain connections in individuals, allowing precise computational models of brain connection patterns to be built. Aim 1. Precision connectome mapping in large-scale human populations. We will develop improved algorithms based on recent methods (Pestilli et al., 2014; Takemura, Caiafa, Wandell and Pestilli 2016) to enable generation of high-resolution whole-brain connectomes from dMRI data, with precise identification of brain connection patterns within human individuals. We propose developing a new generation of connectome mapping methods based on statistical evaluation to study the connectivity without loss in anatomical definition of edges of the connectome (Pestilli et al., 2014; Yeatman et al., 2012). Aim 2. Efficient connectome edge-anatomy recognition. We propose using a new generation of connectomes and statistical validation methods (Pestilli et al. 2014) to implement new methods for connectomics. We will also use recent algorithms for connectome-edge anatomy recognition (Garyfallidis et al. under review; Garyfallidis et al., 2014) to characterize similarity and variability of major white matter tracts across brains in large human populations using data from the Human Connectome Project and the UK-Biobank. Aim 3. Public cloud services to support community-based discovery neuroscience. We plan to leverage the NSF-funded Jetstream-cloud.org public cloud system to deploy a portal with a series of simple web-interfaces and Application Programming Interfaces (APIs) to enable connectome evaluation and mapping methods (presented in Aim 1 and 2) on virtualized computational packages. The web services will be freely shared across the neuroscientific community with the goal of having a broad impact on a diverse communities of engineers, computer scientists and neuroscientists. Aims have different discipline of focus. Aim 1 (computational neuroscience) will develop new computational methods for evaluation of brain connectomes. Aim 2 (computer engineering) will develop algorithms for connectome edge anatomy mapping. Aim 3 (cloud engineering) will bridge cutting-edge scientific cloud technologies to neuroscience and allow reproducible scientific computing Jetstream and virtualized computations.