The Computational Challenges of Constructing and Visualizing Large-Scale Brain Networks

• Activity Number: 166
• Type: Topic Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Computing
• Author(s): Moo Chung*
• Companies: University of Wisconsin - Madison
• Keywords: large-scale brain networks ; diffusion tensor imaging ; MRI ; DTI ; fMRI ; graphs

Abstract

With nearly 100 billion neurons and 100 trillion connections, the human brain remains one of the greatest challenges in science. Diffusion tensor imaging (DTI) technique's ability to visualize such dense connections between different parts of the brain at voxel level has emerged as a major breakthrough. In DTI, tractography algorithms are used to trace up to half million white matter fiber tracts in the brain, which can results in extremely dense networks with up to half million edges and one million nodes. The resulting network defies standard network modeling techniques due to the sheer computational load. The corresponding connectivity matrix contains 250 billion entries, which can't be easily accommodated by desktop computers. Thus, it is critical to reduce the complexity of networks for meaningful biological interpretation and visualization. In this talk, we present computational solutions for overcoming this type of large-scale brain network construction issue in various brain imaging modalities such as MRI, DTI and fMRI. The talk is, in part, based on the study http://arxiv.org/abs/1409.0177.