How to standardize Denoise or Degibbs performance between scans

Hi MRtrix expert,

I’m testing the repeatability of my pre-processing pipeline and noticed that one scan has a slightly lower FA value than the other (same subject) after all the pre-processing. I think this might come from lots of aspects (e.g., the difference between two scan’s prescan, subject motion during the scan), and I’m looking for all the possible reasons that can explain this.

Denoise and Degibbs have been doing a great job in improving the image quality of the FA map (I can see that noise and Gibbs artifacts being removed after applying two command lines, as shown in the attached figures). However, I’m wondering if I’m going to do quantitative analysis and want it to be repeatable between scans on the same subjects or even across different subjects and to avoid the case where Denoise/Degibbs might be doing a better job in one scan than the other (given all other scanning parameters are same). Are there any special parameters that I need to pay attention to make sure the residual noise level is the similar after denoise or degibbs?

Many thanks and all the best,



That’s likely related to the well-documented sensitivity of FA to noise (see e.g. this paper). The application of denoising will clearly have a strong influence here, especially for low SNR data.