Meeting Banner
Abstract #5233

Robust and fast MCMC sampling of diffusion MRI microstructure models

Robbert Leonard Harms1 and Alard Roebroeck1

1Department of Cognitive Neuroscience, Maastricht University, Maastricht, Netherlands

In this work we extend the Maastricht Diffusion Toolbox (MDT) software package with MCMC capabilities, thereby introducing robust and fast sampling of diffusion MRI microstructure models. MDT's object oriented modular design allows arbitrary specification and combination of models, likelihood functions, prior functions and proposal distributions. GPU based computations allow for ~80x faster MCMC sampling; e.g. the 81 volume example NODDI dataset can be sampled with 15000 samples in 1.5 hour, making MDT suitable for probabilistic inference of all implemented models, both dMRI microstructure as quantitative MRI models. The software is open source and freely available at https://github.com/cbclab.

This abstract and the presentation materials are available to members only; a login is required.

Join Here