Skip to content

Analysis tools for sub-millimetre MRI of the cerebral cortex

Notifications You must be signed in to change notification settings

neurospin/highres-cortex

Repository files navigation

highres-cortex

This is a collection of software designed to process 3D images of the cerebral cortex at a sub-millimetre scale, for example high-resolution MRI. In particular, it implements Bok’s equivolumetric depth, which models the depth of cortical layers while compensating for local cortical curvature. If you use this work in an academic publication, please cite the relevant references (see also doc/references.bib):

  • Yann Leprince, Fabrice Poupon, Thierry Delzescaux, Dominique Hasboun, Cyril Poupon, and Denis Rivière. Combined Laplacian-equivolumic model for studying cortical lamination with ultra high field MRI (7 T). 2015 IEEE 12th International Symposium on Biomedical Imaging (ISBI), IEEE, Apr 2015, New York, United States. pp.580-583, DOI: 10.1109/ISBI.2015.7163940. https://hal-cea.archives-ouvertes.fr/cea-01119475
  • Yann Leprince, Clara Fischer, Jean-François Mangin, Benoît Larrat, Sébastien Mériaux, Cyril Poupon, Isabel Reillo, Victor Borrell, Ophélie Foubet, Roberto Toro, and Denis Rivière. Architectonics-informed partition of the cortex at sub-millimetre resolution. 20th Annual Meeting of the Organization for Human Brain Mapping (OHBM), Jun 2014, Hamburg, Germany. 5, pp.951, 2014, F1000Posters. https://hal-cea.archives-ouvertes.fr/cea-01074735

Installation

highres-cortex is released as part of the official BrainVISA containers since March 2021 (BrainVISA 5.0.0). Please refer to https://brainvisa.info/web/download.html for installation instructions.

Graphical user interface

highres-cortex is available as a minimalistic BrainVISA toolbox since BrainVISA 5.1 (January 2023). Simply launch brainvisa and refer to the in-program documentation of the toolbox.

Command-line usage

This package can be used on the command line, here is a short introduction. It is assumed that you are running the Singularity version on a Linux computer, it should work in the same way in the virtual machine (you just need to remove the bv prefix from all commands).

  1. Prepare your input data: the input that is common to to all processes is classif: a voxel-wise tissue classification image in signed 16-bit pixel type, with 0 for exterior voxels (CSF), 100 for cortical gray matter, and 200 for subcortical white matter.

  2. Run the process that you are interested in. The common interface to all processes is the Capsul command-line, which you can call with bv python -m capsul. Use bv python -m capsul --process-help <process_name> to get help for a specific process. Use bv python -m capsul <process_name> [parameter=value ...] to run a process.

    The most important processes are described below:

    • Equivolumetric depth according to Bok’s model can be computed with highres_cortex.capsul.isovolume. The only mandatory input is classif, the output is equivolumetric_depth. You can fine-tune the process with optional parameters, most importantly advection_step_size can be adapted to the spatial resolution and required accuracy. For example:

      bv python -m capsul highres_cortex.capsul.isovolume classif=classif.nii.gz advection_step_size=0.03 equivolumetric_depth=equivolumetric_depth.nii.gz
      
    • Cortical thickness, according to the Laplace model, can be calculated with two different methods:

      • The upwinding method is very fast, and already has sub-pixel accurracy: highres_cortex.capsul.thickness_upw. The only mandatory input is classif, the output is thickness_image.
      • The advection method is slower, but advection_step_size can be tuned for greater accuracy: highres_cortex.capsul.thickness_adv.
    • For parcellating the cortex into volumetric traverses, highres_cortex.capsul.traverses can be used. The only mandatory input is classif, the output is cortical_traverses. The goal_diameter parameter controls the target diameter of merged regions (in millimetres). The advection_step_size parameter is also relevant for this process.

If you have used highres-cortex before the Capsul interface was introduced (beginning of 2018), you may be using the old shell scripts. See examples/scripts/ for equivalent scripts that make use of the Capsul processes.

Contributing

This repository uses pre-commit to ensure that all committed code follows minimal quality standards. Please install it and configure it to run as a pre-commit hook in your local repository (note that this is done automatically by bv_maker):

# Install pre-commit in a virtual environment
python3 -m venv venv/
. venv/bin/activate
pip install pre-commit

pre-commit install  # install the pre-commit hook

Licence

The source code of this work is placed under the CeCILL licence (see LICENCE.CeCILL.txt). This library contains code that is under the GNU LGPL licence (see src/library/cortex_column_region_quality.tcc), as a result, compiled code must be redistributed under the GNU General Public Licence (see LICENCE.GPLv3.txt).

External code used in this repository

  • Code for numerical diagonalization of 3×3 matrices (src/library/cortex_column_region_quality.tcc) is Copyright 2006 Joachim Kopp, under the GNU LGPL v2.1 or later. Reference: Kopp, Joachim. ‘Efficient Numerical Diagonalization of Hermitian 3x3 Matrices’. International Journal of Modern Physics C 19, no. 03 (March 2008): 523–48. arXiv:physics/0610206.