Denoising

Denoising using Local PCA

PCA-based denoising algorithms are effective denoising methods because they explore the redundancy of the multi-dimensional information of diffusion-weighted datasets.

The algorithm to perform the PCA-based denoising involves the following steps:

• First, we estimate the local noise variance at each voxel.

• Then, we apply PCA in local patches around each voxel over the gradient directions.

• Finally, we threshold the eigenvalues based on the local estimate of sigma and then do a PCA reconstruction

We are using stanford_hardi dataset. You can also use your own data.

The workflow for the LPCA denoising requires the paths to the diffusion input file, as well as the b-values and b-vectors files.

You may want to create an output directory to save the denoised data:

mkdir denoise_lpca_output

Run the following command:

dipy_denoise_lpca data/stanford_hardi/HARDI150.nii.gz data/stanford_hardi/HARDI150.bval data/stanford_hardi/HARDI150.bvec --out_dir "denoise_lpca_output"

This command will denoise the diffusion image and save it to the directory denoise_lpca_output.

In case, you did not specify out_dir, the denoised diffusion image will be saved to a file named dwi_lpca.nii.gz by default, located in the input directory also by default.

Note: Depending on the parameters’ values, the effect of the denoising can be subtle or even hardly noticeable, apparent or visible, depending on the choice. Users are encouraged to carefully choose the parameters.

Denoising using Marcenko-Pastur PCA

The PCA-based denoising algorithm exploits the redundancy across the diffusion- weighted images. This algorithm has been shown to provide an optimal compromise between noise suppression and loss of anatomical information for different techniques such as DTI, spherical deconvolution and DKI.

The basic idea behind the PCA-based denoising algorithms is to remove the components of the data that are classified as noise. The Principal Components classification can be performed based on prior noise variance estimates or automatically based on the Marcenko-Pastur distribution. In addition to noise suppression, the PCA algorithm can be used to get the standard deviation of the noise.

We will use the sherbrooke_3shell dataset in DIPY to showcase this denoising method. As with any other workflow in DIPY, you can also use your own data!

Create a directory where to save the denoised image (e.g.: denoise_mppca_output):

mkdir denoise_mppca_output

In order to run the MPPCA denoising method, we need to specify the location of the diffusion data file, followed by the optional arguments. In this case, we will be specifying the patch radius value and output directory.

So, we will run the command as:

dipy_denoise_mppca data/sherbrooke_3shell/HRADI193.nii.gz --patch_radius 10 --out_dir "denoise_mppca_output"

This command will denoise the diffusion image and save it to the specified output directory.

Denoising using NLMEANS

Using the non-local means filter [Coupe08] and [Coupe11], you can denoise 3D or 4D images and boost the SNR of your datasets.

We will use the cfin_multib dataset in DIPY to showcase this denoising method. As with any other workflow in DIPY, you can also use your own data!

In order to run the NLMEANS denoising method, we need to specify the location of the diffusion data file, followed by the optional arguments. In this case, we will be specifying the sigma and patch radius values and output directory.

Create a directory where to save the denoised image (e.g.: denoise_nlmeans_output):

Then, we will run the command as:

dipy_denoise_nlmeans data/cfin_multib/__DTI_AX_ep2d_2_5_iso_33d_20141015095334_4.nii --sigma 2 --patch_radius 2 --out_dir "denoise_nlmeans_output"

The command will denoise the input diffusion volume and write the result to the specified output directory.

Overview of Denoising Methods

Note: Users are recommended to zoom (click on each image) to see the denoising effect.