Harnessing graphics processing units for improved neuroimaging statistics

Eklund, Anders; Villani, Mattias; LaConte, Stephen M.

doi:10.3758/s13415-013-0165-7

Cited by 3 publications

(2 citation statements)

References 97 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By contrast, graphics processing units (GPUs) have a massively parallel structure designed with hundreds of smaller cores optimized to exploit the data level parallelism of certain applications, utilizing simpler instruction sets and distributing them over multiple cores (Eklund et al 2013a;Hernandez-Fernandez et al 2013). This parallelization can accelerate computationally slow processes such as data visualization, stochastic iteration, and Bayesian simulations including probabilistic tractography (Chang et al 2014;Eklund et al 2013a;Eklund et al 2013b;Hernandez-Fernandez et al 2013;Hernandez-Fernandez et al 2019;Lee and Kim 2013;McGraw and Nadar 2007;Sotiropoulos et al 2013). A popular tool in estimating diffusion parameters for whole brain diffusion MRI is available to be run on both CPU or GPU, with GPU algorithm achieving over 100 times speed-up compared to its CPU algorithm (Behrens et al 2007;Behrens et al 2003;Hernandez-Fernandez et al 2013).…”

Section: Introductionmentioning

confidence: 99%

Comparison of CPU and GPU Bayesian Estimates of Fibre Orientations from Diffusion MRI

Kim¹,

Lj²,

Hernández-Fernández

et al. 2019

Preprint

View full text Add to dashboard Cite

Background: The correct estimation of fibre orientations is a crucial step for reconstructing human brain tracts. A popular and extensively used tool for this estimation is Bayesian Estimation of Diffusion Parameters Obtained using Sampling Techniques (bedpostx), which is able to estimate several fibre orientations per voxel (i.e. crossing fibres) using Markov Chain Monte Carlo (MCMC). However, for fitting a model in a whole diffusion MRI dataset, MCMC can take up to a day to complete on a standard CPU. Recently, this algorithm has been ported to run on GPUs, which can accelerate the process, completing the analysis in minutes or hours. However, few studies have looked at whether the results from the CPU and GPU algorithms differ. In this study, we compared CPU and GPU bedpostx outputs by running multiple trials of both algorithms on the same whole brain diffusion data and compared each distribution of output using Kolmogorov-Smirnov tests. Results: We show that distributions of fibre fraction parameters and principal diffusion direction angles from bedpostx and bedpostx_gpu display few statistically significant differences in shape and are localized sparsely throughout the whole brain. Average output differences are small in magnitude compared to underlying uncertainty. Conclusions: Despite small amount of differences in samples created between CPU and GPU bedpostx algorithms, results are comparable given the difference in operation order and library usage between CPU and GPU bedpostx.

show abstract

Section: Introductionmentioning

confidence: 99%

Comparison of CPU and GPU Bayesian Estimates of Fibre Orientations from Diffusion MRI

Kim¹,

Lj²,

Hernández-Fernández

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…The human connectome project (van Essen et al, 2013) 3 , for example, shares high resolution data from a large number of subjects (the goal is 1200), and a single resting state scan results in a dataset of the size 104 × 90 × 72 × 1200. Third, non-parametric methods based on permutation and Bayesian Markov Chain Monte Carlo (MCMC) methods are more frequently being used to improve neuroimaging statistics (da Silva, 2011; Eklund et al, 2012a, 2013b), but suffer from long processing times compared to conventional parametric methods. Some progress toward parallelization has been made in each of the three major packages commonly used in fMRI-based research (SPM, FSL, and AFNI).…”

Section: Introductionmentioning

confidence: 99%

BROCCOLI: Software for fast fMRI analysis on many-core CPUs and GPUs

Eklund

Dufort

Villani

et al. 2014

Front. Neuroinform.

Self Cite

View full text Add to dashboard Cite

Analysis of functional magnetic resonance imaging (fMRI) data is becoming ever more computationally demanding as temporal and spatial resolutions improve, and large, publicly available data sets proliferate. Moreover, methodological improvements in the neuroimaging pipeline, such as non-linear spatial normalization, non-parametric permutation tests and Bayesian Markov Chain Monte Carlo approaches, can dramatically increase the computational burden. Despite these challenges, there do not yet exist any fMRI software packages which leverage inexpensive and powerful graphics processing units (GPUs) to perform these analyses. Here, we therefore present BROCCOLI, a free software package written in OpenCL (Open Computing Language) that can be used for parallel analysis of fMRI data on a large variety of hardware configurations. BROCCOLI has, for example, been tested with an Intel CPU, an Nvidia GPU, and an AMD GPU. These tests show that parallel processing of fMRI data can lead to significantly faster analysis pipelines. This speedup can be achieved on relatively standard hardware, but further, dramatic speed improvements require only a modest investment in GPU hardware. BROCCOLI (running on a GPU) can perform non-linear spatial normalization to a 1 mm3 brain template in 4–6 s, and run a second level permutation test with 10,000 permutations in about a minute. These non-parametric tests are generally more robust than their parametric counterparts, and can also enable more sophisticated analyses by estimating complicated null distributions. Additionally, BROCCOLI includes support for Bayesian first-level fMRI analysis using a Gibbs sampler. The new software is freely available under GNU GPL3 and can be downloaded from github (https://github.com/wanderine/BROCCOLI/).

show abstract

Comparison of CPU and GPU bayesian estimates of fibre orientations from diffusion MRI

et al. 2022

View full text Add to dashboard Cite

Background The correct estimation of fibre orientations is a crucial step for reconstructing human brain tracts. Bayesian Estimation of Diffusion Parameters Obtained using Sampling Techniques (bedpostx) is able to estimate several fibre orientations and their diffusion parameters per voxel using Markov Chain Monte Carlo (MCMC) in a whole brain diffusion MRI data, and it is capable of running on GPUs, achieving speed-up of over 100 times compared to CPUs. However, few studies have looked at whether the results from the CPU and GPU algorithms differ. In this study, we compared CPU and GPU bedpostx outputs by running multiple trials of both algorithms on the same whole brain diffusion data and compared each distribution of output using Kolmogorov-Smirnov tests. Results We show that distributions of fibre fraction parameters and principal diffusion direction angles from bedpostx and bedpostx_gpu display few statistically significant differences in shape and are localized sparsely throughout the whole brain. Average output differences are small in magnitude compared to underlying uncertainty. Conclusions Despite small amount of differences in output between CPU and GPU bedpostx algorithms, results are comparable given the difference in operation order and library usage between CPU and GPU bedpostx.

show abstract

Harnessing graphics processing units for improved neuroimaging statistics

Cited by 3 publications

References 97 publications

Comparison of CPU and GPU Bayesian Estimates of Fibre Orientations from Diffusion MRI

Comparison of CPU and GPU Bayesian Estimates of Fibre Orientations from Diffusion MRI

BROCCOLI: Software for fast fMRI analysis on many-core CPUs and GPUs

Comparison of CPU and GPU bayesian estimates of fibre orientations from diffusion MRI

Contact Info

Product

Resources

About