Anat-SFSeg

Code for our Medical Image Analysis paper "Anat-SFSeg: Anatomically-guided superficial fiber segmentation with point-cloud deep learning" If you find this code useful in your research please cite

Zhang D, Zong F, Zhang Q, et al.
Anat-SFSeg: Anatomically-guided superficial fiber segmentation with point-cloud deep learning[J].
Medical Image Analysis, 2024, 95: 103165.

Setup

The main environment is:

• cuda 11.3
• torch 1.11.0
• nibabel 4.0.2
• whitematteranalysis 0.4.0 (pip install git+https://github.com/SlicerDMRI/whitematteranalysis.git)
• h5py 3.7.0
• vtk 9.2.6

Datasets

The train and example datasets can be downloaded from https://drive.google.com/file/d/13OEzMAidi74Pn8npvkE3_EHjF8OvtjOw/view?usp=drive_link And you can unzip it and add it to the project's path.

Train Model

cd scripts
bash train.sh 

Inference on your data

You need to apply tractography on your diffusion data to obtain the whole brain tractogram. And you need to apply cortical and subcortical parcellation (reconall in FreeSurfer) on your T1w data to obtain the parcellation result (aparc+aseg.mgz). For example, we provide two subjects' tractograms and parcellations in dataset/YourData. Here we use the unscented Kalman filter (UKF) tractography method to obtain the whole brain tractogram named 'wb_fiber.vtk', and the cortical and subcortical parcellation images 'aparc+aseg.nii.gz' are also provided.

cd scripts
bash test.sh 

The results analysis

The fiber bundle category to which the obtained clusters belong are listed in table swm_labels.csv and dwm_labels.csv. Note that the labels in the table is the last number of the resulting bundle (.vtk) file.

References

Thanks to the code of SupWMA, this is the project we rely on. Please cite the following papers for using the code and/or the training data:

Xue T, Zhang F, Zhang C, et al.
Superficial white matter analysis: An efficient point-cloud-based deep learning framework with supervised contrastive learning for consistent tractography parcellation across populations and dMRI acquisitions.
Medical Image Analysis, 2023, 85: 102759.

Zhang, F., Wu, Y., Norton, I., Rathi, Y., Makris, N., O'Donnell, LJ. 
An anatomically curated fiber clustering white matter atlas for consistent white matter tract parcellation across the lifespan. 
NeuroImage, 2018 (179): 429-447

James G Malcolm, Martha E Shenton, Yogesh Rathi.
Neural tractography using an unscented Kalman filter.
In International Conference on Information Processing in Medical Imaging,2019, pp. 126–138.

O'Donnell LJ, Wells III WM, Golby AJ, Westin CF. 
Unbiased groupwise registration of white matter tractography.
In MICCAI, 2012, pp. 123-130.