Mtnormalize vs dwibiascorrect for 4D DWI

I would like to run a bias field correction on my DWI before fitting
[multi-tissue NODDI] (dmipy/example_multi_tissue_noddi.ipynb at master · AthenaEPI/dmipy · GitHub).

In the past I have used your dwibiascorrect tool, but I saw a thread that suggested correcting the DWI by dividing by the bias field output by mtnormalize (i.e., check_norm output). Would you recommend one bias field correction method over the other for 4D DWI? Would it do better for models fit specifically using gray matter and white matter tissue responses?
Thanks so much for your assistance and advice.

Hi Julio,

Personally I use the output of mtnormalise to retrospectively correct the bias field in my DWIs.

It is however the case that many diffusion models are entirely insensitive to global intensity scaling, as they base everything off of the intensity of the b=0 image and bias field correction applies equivalent scaling to both the b=0 and b>0 data, so you may well find that bias field correction has no consequence on the output of that model; it will depend on the internals of how that model is fit, rather than “whether or not it uses gray matter and white matter tissue responses”.

I do not think that whether or not a particular model intended for use does or does not use gray matter and white matter tissue responses influences the method that is recommended for bias field correction. The bias field is what it is, and we want to estimate and correct it as best we can. Whether or not the strengths or limitations of one bias field estimation over and another are more or less consequential for different diffusion models is beyond what has been shown experimentally.