NLSAM datasets repository

Synthetic and in vivo datasets used in the NLSAM paper, for which the main repo can be found here. The synthetic data is based on an earlier version of phantomas.

The data can be downloaded as zip files on the releases page or cloned locally with

git clone https://github.com/samuelstjean/nlsam_data.git

Acknowledgments

If you use the datasets provided therein, please make sure that you quote the following references in any publications:

St-Jean, S., Coupé, P., & Descoteaux, M.,
Non Local Spatial and Angular Matching: Enabling higher spatial resolution diffusion MRI datasets through adaptive denoising.
Medical Image Analysis, 32(2016), 115–130, 2016.

Please also credit (if applicable) the in vivo acquisition in the acknowledgments section of relevant papers as

Datasets were provided (in part) by the Centre d'imagerie moléculaire de Sherbrooke (CIMS)
and the Sherbrooke Connectivity Imaging Lab (SCIL), Université de Sherbrooke, Sherbrooke, Québec, Canada.

Synthetic data description

The phantomas b1000 and phantomas b3000 folders contains the raw datasets that were used. The masks folders contains everything needed for running a tractometer comparison in the proper folders. The naming convention goes as follow

• hardi-scheme contains the bvals/bvecs used by the phantom, which is the same 64 gradient directions for both diffusion weighting.
• dwis.nii.gz is the noiseless ground-truth data to which noise was added afterward.
• SNR-(10, 15, 20 or 30) is the SNR of the dataset computed as SNR = mean(b0) / sigma, with mean(b0) computed inside white matter (see wm.nii.gz inside the masks folder) and sigma the noise standard deviation.
• coils-(1, 4, 8 or 12) defines the noise distribution as outlined in Eq. 6 of the paper.
• var-3 means that this dataset has a varying noise profile, which is SNR at the edges and SNR/3 near the center in a linear scale.

The datasets I used were (at both b-1000 and b-3000):

• Stationary noise
  • dwis_SNR-10_coils-1.nii.gz
  • dwis_SNR-10_coils-12.nii.gz
  • dwis_SNR-20_coils-1.nii.gz
  • dwis_SNR-20_coils-12.nii.gz
• Spatially variable noise
  • dwis_SNR-15_coils-1_var-3.nii.gz
  • dwis_SNR-15_coils-12_var-3.nii.gz
  • dwis_SNR-20_coils-1_var-3.nii.gz
  • dwis_SNR-20_coils-12_var-3.nii.gz

Note that not all datasets were used in the original paper, but are still provided here for interested users.

In vivo data description

Two diffusion weighted scans and a T1 weighted scan of a healthy volunteer were acquired on a 3T Philips scanner. A SENSE acceleration factor R = 2 (producing spatially varying Rician noise) was used with a gradient strength of 45 mT/m and a 32 channels head coil. The following acquisition parameters were used :

For the high resolution dataset :

• 40 gradient directions at b = 1000 s/mm² + 1 b0 image at 1.2 mm voxel size
• TR/TE = 18.9 s / 104 ms
• Total acquisition time : 13 mins

For the standard resolution dataset :

• 64 gradient directions at b = 1000 s/mm² + 1 b0 image at 1.8 mm voxel size
• TR/TE = 11.1 s / 63 ms
• Total acquisition time : 12 mins

License

All content is available under the Creative Common Attribution license, see LICENSE for more information.