Patch Denoising Methods#

LICENSE

This repository implements patch-denoising methods, with a particular focus on local-low rank methods.

The target application is functional MRI thermal noise removal, but this methods can be applied to a wide range of image modalities.

It includes several local-low-rank based denoising methods (see the documentation for more details):

MP-PCA
Hybrid-PCA
NORDIC
Optimal Thresholding
Raw Singular Value Thresholding

A mathematical description of these methods is available in the documentation.

Installation#

$ pip install patch-denoise

patch-denoise requires Python>=3.9

Quickstart#

After installing you can use the patch-denoise command-line.

$ patch-denoise input_file.nii output_file.nii --mask="auto"

See patch-denoise --help for detailed options.

Documentation and Examples#

Documentation and examples are available at https://paquiteau.github.io/patch-denoising/

Development version#

$ git clone https://github.com/paquiteau/patch-denoising
$ pip install -e patch-denoising[dev,doc,test,optional]

Citation#

If you use this package for academic work, please cite the associated publication, available on HAL

@inproceedings{comby2023,
  TITLE = {{Denoising of fMRI volumes using local low rank methods}},
  AUTHOR = {Pierre-Antoine, Comby and Zaineb, Amor and Alexandre, Vignaud and Philippe, Ciuciu},
  URL = {https://hal.science/hal-03895194},
  BOOKTITLE = {{ISBI 2023 - International Symposium on Biomedical Imaging 2023}},
  ADDRESS = {Carthagena de India, Colombia},
  YEAR = {2023},
  MONTH = Apr,
  KEYWORDS = {functional MRI ; patch denoising ; singular value thresholding ; functional MRI patch denoising singular value thresholding},
  PDF = {https://hal.science/hal-03895194/file/isbi2023_denoise.pdf},
  HAL_ID = {hal-03895194},
  HAL_VERSION = {v1},
}