freesurfer / samseg

This repository contains the code for the SAMSEG.

SAMSEG

This repository is under construction. Please look at the main repository in the meantime.

TODO: Add segmentation image(s)?

Sequence Adaptive Multimodal SEGmentation (SAMSEG) is a tool to robustly segment dozens of brain structures from head MRI scans without preprocessing. The characteristic property of SAMSEG is that it accepts multi-contrast MRI data without prior assumptions on the specific type of scanner or pulse sequences used. Dedicated versions to handle longitudinal data, or to segment white matter lesions in multiple sclerosis (MS) patients are also available.

TODO: The description above does not include the subregions module. Fix this.

Build Status

Linux	Windows	MacOS

Getting Started

SAMSEG runs on 64bit Windows, Linux, and MacOS machines. Please visit the official SAMSEG Wiki and subregions Wiki for instructions.

Most of the functionalities of SAMSEG do not require FreeSurfer to be installed on your system, except:

• longitudinal registration preprocessing;
• subregions module.

Installing from source (on *nix)

1. Clone project: git clone https://github.com/freesurfer/samseg.git

2. Get the submodules: git submodule init git submodule update

3. Create a virtual environment using, e.g., conda: conda create -n samseg python=3.9

4. Activate the virtual environment: conda activate samseg

5. Install correct compilers for ITK v.4.13.2 conda install -c conda-forge gxx_linux-64=7.5 gcc_linux-64=7.5 sysroot_linux-64=2.17

6. Create the ITK build directory mkdir ITK-build cd ITK-build

7. Export compilers installed with conda: export CC=<your_conda_path>/envs/samseg/bin/x86_64-conda_cos6-linux-gnu-gcc export CXX=<your_conda_path>/envs/samseg/bin/x86_64-conda_cos6-linux-gnu-g++

8. Run CMAKE: cmake -DBUILD_SHARED_LIBS=OFF -DBUILD_TESTING=OFF -DBUILD_EXAMPLES=OFF -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=../ITK-install ../ITK

9. Install: make install cd..

10. Install in development mode (.[test] installs pytest, use just . if you don't want to run tests) ITK_DIR=ITK-install python -m pip install --editable .[test]

11. If you want to build wheels call ITK_DIR=ITK-install python -m pip wheel . -w ./dist --no-deps

12. If you're developing and want to run the tests call (in the root of the git repository): pytest samseg

References

If you use these tools in your analysis, please cite:

• Cross-sectional: Fast and sequence-adaptive whole-brain segmentation using parametric Bayesian modeling. O. Puonti, J.E. Iglesias, K. Van Leemput. NeuroImage, 143, 235-249, 2016.

• MS lesions: A Contrast-Adaptive Method for Simultaneous Whole-Brain and Lesion Segmentation in Multiple Sclerosis. S. Cerri, O. Puonti, D.S. Meier, J. Wuerfel, M. Mühlau, H.R. Siebner, K. Van Leemput. NeuroImage, 225, 117471, 2021.

• Longitudinal: An Open-Source Tool for Longitudinal Whole-Brain and White Matter Lesion Segmentation. S. Cerri, D.N. Greve, A. Hoopes, H. Lundell, H.R. Siebner, M. Mühlau, K. Van Leemput. NeuroImage: Clinical, 38, 103354, 2023.

• Thalamus: A probabilistic atlas of the human thalamic nuclei combining ex vivo MRI and histology. Iglesias, J.E., Insausti, R., Lerma-Usabiaga, G., Bocchetta, M., Van Leemput, K., Greve, D., van der Kouwe, A., Caballero-Gaudes, C., Paz-Alonso, P. Neuroimage (accepted).

• Hippocampus: A computational atlas of the hippocampal formation using ex vivo, ultra-high resolution MRI: Application to adaptive segmentation of in vivo MRI. Iglesias, J.E., Augustinack, J.C., Nguyen, K., Player, C.M., Player, A., Wright, M., Roy, N., Frosch, M.P., Mc Kee, A.C., Wald, L.L., Fischl, B., and Van Leemput, K. Neuroimage, 115, July 2015, 117-137.

• Amygdala: High-resolution magnetic resonance imaging reveals nuclei of the human amygdala: manual segmentation to automatic atlas. Saygin ZM & Kliemann D (joint 1st authors), Iglesias JE, van der Kouwe AJW, Boyd E, Reuter M, Stevens A, Van Leemput K, Mc Kee A, Frosch MP, Fischl B, Augustinack JC. Neuroimage, 155, July 2017, 370-382.

• Brainstem: Bayesian segmentation of brainstem structures in MRI. Iglesias, J.E., Van Leemput, K., Bhatt, P., Casillas, C., Dutt, S., Schuff, N., Truran-Sacrey, D., Boxer, A., and Fischl, B. NeuroImage, 113, June 2015, 184-195.

• Longitudinal subregions: Bayesian longitudinal segmentation of hippocampal substructures in brain MRI using subject-specific atlases. Iglesias JE, Van Leemput K, Augustinack J, Insausti R, Fischl B, Reuter M. Neuroimage, 141, November 2016, 542-555.

samseg's Issues

how to use the mesh created by samseg?

I put --save-mesh at samseg command, and finally samseg created a mesh file. but it can not be used in freeview as rh.pial (recon-all). So, how to use the mesh files created by samseg?

is there any statistical concern on removing the gaussian probability from timepoint model

hello dr cerri, first of all, thank you for your great work in samseg. i noticed you are removing the gaussian probability from timepoint model in the longitudinal fitmodel. is there any specific statistical concern.

Optimsing SAMSEG

Hello!
I am a masters student trying to optmise SAMSEG to detect WMH in stroke. I was wondering if it would be possible to add stroke images (with manually segmented WMH) to optimise the hyperparameters for modelling spatial location and shape of lesions. If so, I am struggling to find how to adapt the code to add images from or dataset for optimisation proposes, as I am new to coding.
Kind regards,
Adrija.

Does Samseg use the GPU?

Dear Developers,

without cuda drivers, samseg issues a warning at the beginning that the GPU will not be used. This warning disappears after the cuda drivers are installed but the execution time was the same as before and "watch nvidia-smi" does not show any GPU usage at all.
Does the current version use the GPU at all? If not, do you have a GPU compiled version available or could you provide a short guide how to compile samseg with GPU support?

Thanks a lot!

Kind regards,
Lukas

How do I train samseg with my own data

In practical application, the segmentation effectiveness of a single MRI structure is not consistently robust in certain areas. Therefore, I intend to train a new probabilistic model tailored to my own data.
Thanks.