Accelerating Diffusion Tensor Estimation using General-Purpose Graphics Processing Unit

Lin-Ching Chang¹, Mikhail a Gorbachev¹

A general-purpose graphics processing unit (GPU) offers a powerful processing platform to accelerate non-graphics applications such as tensor estimation in Diffusion Tensor Imaging (DTI). Diffusion tensor maps are computed on a voxel-by-voxel basis by fitting the signal intensities of diffusion weighted images as a function of their corresponding b-matrices. This computation can be significantly accelerated by using the GPU. This study presents the application of using GPU hardware in diffusion tensor estimation by accelerating the weighted multivariate linear regression. The results are tested in simulated 3D brain dataset and show faster computation time against the CPU. The proposed GPU framework can accelerate DTI simulation and can be readily applied to quantitative assessment of the DTI using bootstrap analysis.