Blue matter

Think Global, Act Local; Projectome Estimation with BlueMatter
Anthony J. Sherbondy1 , Robert F. Dougherty2 , Rajagopal Ananthanarayanan1, Dharmendra S. Modha1 , and Brian A. Wandell2
1 2

IBM Almaden Reserach Center, Almaden, USA anthony.sherbondy@stanford.edu Psychology Department, Stanford University, USA

Abstract. Estimating the complete set of white matter fascicles (the projectome)from diffusion data requires evaluating an enormous number of potential pathways; consequently, most algorithms use computationally efficient greedy methods to search for pathways. The limitation of this approach is that critical global parameters - such as data prediction error and white matter volume conservation - are not taken into account. We describe BlueMatter, a parallel algorithm for globalprojectome evaluation, which uniquely accounts for global prediction error and volume conservation. Leveraging the BlueGene/L supercomputing architecture, BlueMatter explores a massive database of 180 billion candidate fascicles. The candidates are derived from several sources, including atlases and mutliple tractography algorithms. Using BlueMatter we created the highest resolution,volume-conserved projectome of the human brain.

1

Introduction

The white matter of the human brain comprises more than 150km of long-range myelinated connections [1]. Understanding the architecture of these long-range projections (the projectome) is important for understanding brain function [2]. Diffusion weighted imaging fiber tractography (DWI-FT) is the only non-invasive method for studying the humanbrain projectome. While there has been great progress in developing fiber tracking techniques [3,4,5], there is wide agreement that current methods fail in many specific cases [6,7,4]. A limitation is that current algorithms find pathways using greedy techniques; that is, the algorithms make decisions based on individual tracts without considering the entire projectome. Further, current algorithmsdo not optimize
Sponsored by Defense Advanced Research Projects Agency, Defense Sciences Office (DSO), Program: Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE), Issued by DARPA/CMO under Contract No. HR0011-09-C-0002. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, eitherexpressly or implied, of the Defense Advanced Research Projects Agency or the U.S. Government. Also supported by NEI EY01500.
G.-Z. Yang et al. (Eds.): MICCAI 2009, Part I, LNCS 5761, pp. 861–868, 2009. c Springer-Verlag Berlin Heidelberg 2009

862

A.J. Sherbondy et al.

the projectome to fit the original diffusion data. Finally, they ignore important physical constraints, such as the volumeoccupied by the estimated fascicles. We introduce the BlueMatter algorithm to address these limitations. BlueMatter takes as input fascicles derived from multiple tractography algorithms. It searches for an optimal combination of these fascicles (the projectome) subject to two error terms. The first term compares the predicted and measured diffusion-weighted images. We refer to this constraint asdiffusion-fitting (see also [8,9]). The second term accounts for an upper limit on the fascicle volume. We refer to this constraint as volume regularization; this term is an important physical constraint that helps resolve the ill-posed inverse problem of finding a projectome that minimizes diffusion-fitting error. Accounting for diffusion-fitting and fiber count was discussed by Zhang and Laidlaw [10]; however,their technique did not address fascicle volume estimation and was limited to fascicles derived from deterministic algorithms. BlueMatter is the first algorithm to produce a human brain projectome that combines diffusion-fitting and volume regularization. BlueMatter combines fascicle estimates from deterministic and stochastic DWI-FT algorithms; uniquely integrating algorithms with different...