Skip to content

MRtrix3 DWI Denoising (MP-PCA)

Library: MRtrix3 | Docker Image: mrtrix3/mrtrix3

Function

Removes thermal noise from diffusion MRI data using Marchenko-Pastur PCA exploiting data redundancy across diffusion directions.

Modality: 4D diffusion-weighted NIfTI with multiple diffusion directions (minimum ~30 recommended).

Typical Use: First step in DWI preprocessing to improve SNR.

Key Parameters

, -noise (output noise map), -extent (spatial patch size, default 5,5,5), -mask (brain mask)

Key Points

Should be run FIRST, before any other processing. Requires sufficient number of diffusion directions. Noise map useful for QC.

Inputs

Name Type Required Label Flag
input File Yes Input DWI image
output string Yes Output denoised image
noise string No Output noise map filename -noise
extent string No Sliding window extent (e.g., 5,5,5) -extent
mask File No Processing mask -mask

Accepted Input Extensions

  • input: .mif, .nii, .nii.gz, .mgh, .mgz
  • mask: .mif, .nii, .nii.gz, .mgh, .mgz

Outputs

Name Type Glob Pattern
denoised File $(inputs.output)
noise_map File `${
if (inputs.noise)
else
}
`
log File dwidenoise.log

Output Extensions

  • denoised: .mif, .nii, .nii.gz
  • noise_map: .mif, .nii, .nii.gz

Docker Tags

Available versions: latest, 3.0.8, 3.0.7, 3.0.5, 3.0.4, 3.0.3

Categories

  • Diffusion MRI > MRtrix3 > Preprocessing

Documentation

Official Documentation