Meeting Banner
Abstract #2626

Learning microstructure parameters from diffusion-weighted MRI using random forests

Gemma Nedjati-Gilani 1 , Matt G Hall 1 , Claudia Angela M Wheeler-Kingshott 2 , and Daniel C Alexander 1

1 Computer Science & Centre for Medical Image Computing, University College London, London, London, United Kingdom, 2 Department of Neuroinflammation, Institute of Neurology, UCL, London, United Kingdom

Deriving analytical models of the diffusion MR signal which account for permeability is inherently difficult and often requires strong assumptions to be made about the compartmentation of water within the tissue. Given these problems, in this study we construct a computational model using Monte Carlo simulations and machine learning. We use random forest regression to learn a mapping between simulations and microstructure parameters and obtain an efficient and accurate model for microstructure imaging that accounts for permeability. We show that unseen microstructure parameters are well-predicted by the random forest regressor for both noise-free and noisy data.

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

Join Here