Rician noise

1

Dear experts,

This question may sound a bit left field…

Does dwibiascorrect (with ANTS N4) correct for rician noise?

If not, any advice on how to perform rician noise correction using MRtrix3?

Cheers,

Joseph

2

Hi Joseph,
No, the dwibiascorrect script only performs intensity inhomogeneity correction (i.e. bias field correction). However, I can appreciate the name is a little ambiguous.

@dchristiaens recently implemented this denoising method (see the dwidenoise command). In my experience the denoising works really well, and unlike other denoising methods you don’t have to worry about it over smoothing and blurring anatomy.

In terms of a Rician bias correction, the output noise (sigma) maps from dwidenoise can be used to correct the Rician bias, and this is something we hope to add to MRtrix in the future. Unfortunately it’s not a trivial addition so I’m not sure when this will be available.
Cheers,
Dave

3

Hello, Dave,

Thanks for the response re Rician noise.

In fact, I have also been looking into methods for correcting Gibbs-Ringing artifacts without significantly compromise image resolution by over smoothing.

I note G-R is a common artifact for our MBMS DWI data.

Contrary to your experience, I found “dwidenoise” doesn’t do a good job correcting for G-R.

I have experimented another method by Kellner et al (http://dx.doi.org/10.1002/mrm.26054), which appeared to perform better than Christlaen’s method, correcting for G-R.

Joseph

4

Hi Joseph,
We plan to add a separate command for Gibbs ringing removal using this approach.

@bjeurissen has already implemented it in MRtrix, however we still just need to remove a dependency on an external library (fftw) before making it public. Hopefully one of the dev team will have time to do this at some point soon. Will post back here when it is finished.
Cheers,
Dave

5

Thanks for the information, Dave.

Cheers,

Joseph