microTorch: A software package for fast and flexible self-supervised diffusion MRI model fitting

Sen, Snigdha, Ahmed, Rajib, Arends, Gerrit C., Casali, Nicola, Gomez, Alvaro Planchuelo, Chen, Xiaoxiang, Masramon, Marta, Parker, Christopher S., Palombo, Marco ORCID: https://orcid.org/0000-0003-4892-7967, Tax, Chantal M. W., Panagiotaki, Eleftheria and Slator, Paddy J. ORCID: https://orcid.org/0000-0001-6967-989X 2026. microTorch: A software package for fast and flexible self-supervised diffusion MRI model fitting. Journal of Open Research Software 14 (1) , 59. 10.5334/jors.736

PDF - Published Version Available under License Creative Commons Attribution. Download (1MB)

Abstract

The microTorch software package allows for the fast and flexible fitting of a wide range of multi-compartment microstructure models to diffusion magnetic resonance imaging (dMRI) data. It employs a self-supervised learning approach, which enables the speed gains of deep learning whilst minimising estimation bias induced by the training data distribution. Its modular architecture enables users to readily evaluate different combinations of compartment models, apply established multi-compartment models, and interchange neural network architectures without modifying the underlying framework. The package supports custom acquisition schemes and provides configurable training settings, allowing it to be tailored to specific datasets. microTorch is built using PyTorch and is freely available at https://github.com/snigdha-sen/microtorch.

Item Type:	Article
Date Type:	Published Online
Status:	Published
Schools:	Schools > Computational & Mathematical Sciences Schools > Psychology Schools > Computer Science & Informatics
Publisher:	Ubiquity Press
ISSN:	2049-9647
Date of First Compliant Deposit:	7 September 2026
Date of Acceptance:	13 August 2026
Last Modified:	07 Sep 2026 14:15
URI:	https://orca.cardiff.ac.uk/id/eprint/189441

Actions (repository staff only)

Edit Item