Cited by 2,974Advanced Neuroimaging Techniques and ApplicationsAdvanced MRI Techniques and ApplicationsFunctional Brain Connectivity Studies

The DOI matches its Crossref record; 59 of the 59 references it cites were found in a registry or an index; 5 of the 9 sources its claims cite were checked for support.

MRtrix3: A fast, flexible and open software framework for medical image processing and visualisation

Open access

Published version of a preprint

Abstract

MRtrix3 is an open-source, cross-platform software package for medical image processing, analysis and visualisation, with a particular emphasis on the investigation of the brain using diffusion MRI. It is implemented using a fast, modular and flexible general-purpose code framework for image data access and manipulation, enabling efficient development of new applications, whilst retaining high computational performance and a consistent command-line interface between applications. In this article, we provide a high-level overview of the features of the MRtrix3 framework and general-purpose image processing applications provided with the software.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

MRtrix3 is an open-source, cross-platform C++11/OpenGL software framework for medical image processing, analysis and visualisation, with a particular emphasis on diffusion MRI. The paper is a high-level overview of its architecture, guiding principles, viewer (MRView), bundled command-line tools, and a benchmark of a typical multi-shell dMRI pipeline.

Key findings

  • MRtrix3 provides a modular, general-purpose image access and multi-threading framework with a consistent command-line interface; it is released under the Mozilla Public Licence 2.0, version-controlled with Git, and includes continuous integration testing.
  • The framework supports native reading/writing of DICOM (read-only), Analyze, NIfTI1, NIfTI2, MGH and its own formats, with arbitrary data strides (only its own formats support arbitrary stride ordering) and automatic history tracking in output file headers.
  • Bundled tools span basic image operations (mrinfo, mrconvert, mrcat, mrcalc, mrmath, mrstats) and dMRI analyses including MP-PCA denoising, Gibbs ringing removal, DTI, single- and multi-tissue CSD, deterministic/probabilistic/global tractography, ACT, SIFT/SIFT2, and fixel-based analysis.
  • Benchmark: a full pipeline from raw DICOM through pre-processing to 1 million probabilistic streamlines took 52 minutes total on an Intel i7-4930K (6 cores, 3.40 GHz), 32 GB RAM, NVIDIA GTX 780; the largest single step was distortion/eddy correction via FSL (32m59s), followed by tckgen (7m33s) and 5ttgen (7m29s).

Methods

  • Design/description paper: the authors present the guiding principles, software architecture (image access, strides, DICOM handling, Unix pipes, ThreadedLoop and run_queue multi-threading APIs, build system, Python scripting library) and the MRView visualisation tool.
  • Runtime benchmark used the B.A.T.M.A.N. tutorial dataset: Siemens Prisma 3T, 64-channel head coil, structural T1 (1 mm isotropic, 160 slices, 5 min) plus multi-shell dMRI (2.5 mm isotropic, b = 0/1000/2000/3000 s/mm² with 15/17/31/50 volumes, 17 min acquisition), processed on the desktop system described above.
  • No formal validation study, accuracy comparison, or user study is reported; the paper is a software overview with an illustrative timing table.

Limitations

  • The paper states no explicit limitations; it instead lists future directions (continued high-performance method implementations, broader adoption and tutorials, and encouraging C++ library use beyond diffusion MRI).
  • Evident from the text: the runtime benchmark is a single dataset on a single hardware configuration, so timings are illustrative rather than a general performance characterisation; no accuracy or correctness evaluation against alternative packages is presented; and the software's historical focus remains diffusion MRI despite the general-purpose API.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1016/j.neuroimage.2019.116137

Cites

59 of 59 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

74 No basis given · 3 Own method · 3 Own interpretation · 3 Own result · 2 Supported · 2 Partly supported · 1 Sources not found · 1 No full text of the sources · 1 Not checked yet · accuracy not yet measured

Show all 16 claims

  • It may state two separate claims as one. 24
  • Probably not a substantial claim. 64
  • The quote only partly states this claim. 1
74 of 74
  • The authors state this as an assumption.

    Reproducible neuroscience relies first and foremost on open source software.

    “The MRtrix3 collective believes that reproducible neuroscience relies first and foremost on open source software.”
  • The paper relies on this without support.

    The multi-shell diffusion MRI dataset used for the runtime breakdown is typical of routine clinical practice.

    “The following provides a breakdown of the time required to process a typical multi-shell diffusion MRI dataset as might be acquired in routine clinical practice.”
2 of 2

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Research works printed but not linked: 1

  1. Raffelt, 2017, Bias field correction and intensity normalisation for quantitative analysis of apparent fibre density, Proc. Intl. Soc. Mag. Reson. Med., 3541

Cited by

Sorted by citations

See all 2,974

Metadata

Authors
Tournier, J.-D., Smith, R., Raffelt, D., Tabbara, R., Dhollander, T., Pietsch, M., Christiaens, D., Jeurissen, B., Yeh, C.-H., & Connelly, A.
Affiliations
  • King's College London (GB)
  • The University of Melbourne (AU)
  • Florey Institute of Neuroscience and Mental Health (AU)
  • University of Antwerp (BE)
Corresponding author
Tournier, J-Donald
Venue
NeuroImage
Publisher
Elsevier BV
Published
2019-11
Volume
202
Pages
116137
ISSN
1053-8119 (ISSN-L)
OpenAlex
W2970898057
Cited by
2,974
References
57
Citations by year
2026: 495 · 2025: 627 · 2024: 491 · 2023: 432 · 2022: 408 · 2021: 331 · 2020: 171 · 2019: 18
FWCI
139.11
Type
journal article
Language
en
Open access
Locations
  • NeuroImage (publishedVersion)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
  • Lirias (submittedVersion, other-oa)
Funders
  • King's College London
  • Seventh Framework Programme (319456)
  • European Research Council
  • National Imaging Facility
  • National Health and Medical Research Council (1026077, 1091593, 400121, 628952)
  • Medical Research Council (MR/K006355/1)
  • Fonds Wetenschappelijk Onderzoek (12M3116N, 12M3119N)
  • Engineering and Physical Sciences Research Council
  • State Government of Victoria
  • National Institute for Health and Care Research
  • Wellcome Trust
  • Guy's and St Thomas' NHS Foundation Trust
  • King’s College London (WT 203148/Z/16/Z)
Topics
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Advanced Neuroimaging Techniques and Applications
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Advanced MRI Techniques and Applications
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Functional Brain Connectivity Studies
Keywords
MRtrix3, diffusion MRI, neuroimaging, medical image processing, image visualization, open source software, command-line interface, software framework
MeSH terms
Humans, Image Processing, Computer-Assisted, Magnetic Resonance Imaging, Software Design, Diffusion Magnetic Resonance Imaging, Neuroimaging

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-10-24, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-24 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review