dmriprep
dmriprep copied to clipboard
Roundup of denoising and unringing algorithms from ISMRM
Patch2Self was brought to my attention by a colleague and was also announced on Neurostars yesterday: https://neurostars.org/t/patch2self-self-supervised-denoiser-for-diffusion-mri-data-let-your-old-data-live-a-new-life/17671
The paper explaining more can also be found here: https://arxiv.org/pdf/2011.01355.pdf They include comparisons with MP-PCA.
The method is also implemented in dipy.
EDIT: I'm editing this issue to make note of some of the new methods and comparisons introduced at ISMRM.
Looks incredible. Although, I have to say that something presented as "magic" makes me nervous.
Denoising
- MP-PCA: implemented in dipy, MRtrix and DESIGNER
- MP-PCA non-local patch:
- Patch2Self: implemented in dipy
- using NORDIC in combination with Patch2Self?
Unringing
- subvoxel shifts
- RPG correction: implemented in DESIGNER
Hi @josephmje and @oesteban ! I only saw this now :) I am happy to help with Patch2Self
if this is still of interest! Also nothing magical in Patch2Self really...
Let me know what I can help with! I am also working on NORDIC + Patch2Self
--> Do you guys have a Pythonic Implementation of NORDIC
? Here is an example of the excellent potential it has: