Using GPUs to accelerate computational diffusion MRI: From microstructure estimation to tractography and connectomes

Abstract

The great potential of computational diffusion MRI (dMRI) relies on indirect inference of tissue microstructure and brain connections, since modelling and tractography frameworks map diffusion measurements to neuroanatomical features. This mapping however can be computationally highly expensive, particularly given the trend of increasing dataset sizes and the complexity in biophysical modelling. Limitations on computing resources can restrict data exploration and methodology development. A step forward is to take advantage of the computational power offered by recent parallel computing architectures, especially Graphics Processing Units (GPUs). GPUs are massive parallel processors that offer trillions of floating point operations per second, and have made possible the solution of computationally-intensive scientific problems that were intractable before. However, they are not inherently suited for all problems. Here, we present two different frameworks for accelerating dMRI computations using GPUs that cover the most typical dMRI applications: a framework for performing biophysical modelling and microstructure estimation, and a second framework for performing tractography and long-range connectivity estimation. The former provides a front-end and automatically generates a GPU executable file from a user-specified biophysical model, allowing accelerated non-linear model fitting in both deterministic and stochastic ways (Bayesian inference). The latter performs probabilistic tractography, it can generate whole-brain connectomes and supports new functionality for imposing anatomical constraints, such as inherent consideration of surface meshes (GIFTI files) along with volumetric images. We validate the frameworks against well-established CPU-based implementations and we show that despite the very different challenges for parallelising these problems, a single GPU achieves better performances than 200 CPU cores thanks to our parallel designs.

Estimating axial diffusivity in the NODDI model

“…Note that the absolute limit of the axial diffusivity in pure water at 37°C is ~ 3 — 3.1 μm 2 /ms, far less than the upper limit of 4 μm 2 /ms used here. To reduce computation time, the in vivo data were optimised using the analytic solution and graphical processing units (GPUs) using FSL’s CUDA Diffusion Modelling Toolbox, cuDIMOT [24]. The model was optimised using Markov Chain Monte Carlo (MCMC) - which afforded estimation of each parameter’s distribution - and Rician noise modelling, both of which are internal functions of cuDIMOT [24].…”

XTRACT - Standardised protocols for automated tractography in the human and macaque brain

“…The full list of tracts that are currently supported is presented in Table 1. We further implemented a new cross-species tractography (XTRACT) toolbox, capable of reading the standard space tractography protocols and performing probabilistic tractography (Behrens et al, 2007), with the option of GPU acceleration (Hernandez-Fernandez et al, 2019). Figure 1 illustrates the main stages for a single tractography protocol.…”

Modelling white matter in gyral blades as a continuous vector field

“…Group-average analysis were carried out on datasets aligned using MSMAll intersubject registration (Robinson et al 2014;Glasser et al 2016). We used FSL's probtrackx2 (Behrens et al 2007;Hernandez-Fernandez et al 2019) to compare the features of the connectome when seeding/terminating streamlines at either the white/grey-matter boundary or at the new interface between the gyral and deep white matter. For these two surfaces we (i) compared the density distribution of streamline endpoints when seeding from the subcortical volume or from the contralateral hemisphere, (ii) assessed the similarity in the path that streamlines seeded from the surface take through deep white matter, and (iii) performed a comparison between the functional and structural connectome.…”