DWI-Denoising-Anatomical info

Hello all,
I am a new user in MRtrix. I need to denoise my DWI image in a preprocessing stage. My DWI data is acquired with 2 b values, 0, and 1000. I used dwidenoise command to denoise my original image, and used mrcalc to calculate the residual. However, I get anatomical information in my residual.mif. I tried increasing the denoising extent to even 11. However, I keep getting pretty much the same result. I’d appreciate any guidnace you could provide.


It would help if you could you post a screenshot of the noise and residual maps that you get, so we can see what that structure looks like.



Hi Donald,
Thanks for your prompt response. Attached please find the images. The first image belongs to residual map for b=0.

This image is a corresponding noise map for b=0.

That’s a map of the residuals…? Wow, that’s definitely not what I’d expect to see…

Can you post the exact commands that you used to generate these images, starting with the dwidenoise call?

Hi Donald,

I used the following commands for default extension:
dwidenoise WS-Feb.mif WS-Feb-den.mif -noise noise.mif
mrcalc WS-Feb.mif WS-Feb-den.mif -subtract residual.mif
mrview residual.mif

I also tried other extensions, such as 7, 19. None of them worked.

Thank you,

Nothing wrong there… Can you post the output of:

mrinfo WS-Feb.mif
dirstat WS-Feb.mif

And maybe @dchristiaens has some idea what might be going on…?

HI Donald,

The following is the result of mrinfo command:

Image name: “WS-Feb.mif”

Dimensions: 192 x 192 x 28 x 2
Voxel size: 1.25 x 1.25 x 6 x 6
Data strides: [ -1 2 3 4 ]
Format: MRtrix
Data type: signed 16 bit integer (little endian)
Intensity scaling: offset = 0, multiplier = 1
Transform: 0.9955 0.0574 -0.07533 -115.9
-0.06037 0.9975 -0.03772 -112.3
0.07297 0.0421 0.9964 -176.9
command_history: mrconvert AX_DWI_WS_Feb2020_AX_DWI_20200212141245_5.nii WS-Feb.mif -fslgrad AX_DWI_WS_Feb2020_AX_DWI_20200212141245_5.bvec AX_DWI_WS_Feb2020_AX_DWI_20200212141245_5.bval (version=3.0.2)
comments: TE=1e+02;Time=142747.165;phase=1
dw_scheme: 0,0,0,0
mrtrix_version: 3.0.2

result of dirstat command:

dirstat WS-Feb.mif
dirstat: [WARNING] The following image volumes were not successfully assigned to a b-value shell:
dirstat: [WARNING] 1 (1000)
WS-Feb.mif (b=0) [ 1 directions ]

Bipolar electrostatic repulsion model:
nearest-neighbour angles: mean = 180, range [ 180 - 180 ]
energy: total = 0, mean = 0, range [ 0 - 0 ]

Unipolar electrostatic repulsion model:
nearest-neighbour angles: mean = 180, range [ 180 - 180 ]
energy: total = 0, mean = 0, range [ 0 - 0 ]

Spherical Harmonic fit:
condition numbers for lmax = 2 → 0: [ ]

Asymmetry of sampling:
norm of mean direction vector = 0

1 Like

Ah, OK, there’s only 2 volumes in there. That’s not enough to do denoising using this approach, there’s not enough redundancy. It’s designed to operate on a full DWI series, so unfortunately can’t be used like this.

I assumed so. Thank you, Donald. I just wanted to make sure I haven’t done anything wrong here. I will try this on full DWI series, soon.