Kurtosis parameters estimation

Hi all MRtrixers,
after I preprocessed my data using MRtrix3, in some steps FSL respectively (b values 700, 1000 and 2300) I need to calculate kurtosis parameters. I found 2 ways for it - DESINGER or DKE (Diffusional Kurtosis Estimator). Is there any suggestion/recommendation in using one of theme? Or your personal opinion?
I saw J Varaart, 2013 in references for both dwi2tensor and Designer’s dki_fit function on github. Does it mean that I can calculate kurtosis tensor using dwi2tensor and for calculation of metrics I can call just Designer’s dki_parameters function?

I found that DESIGNER use some of MRtrix3 functions. That this mean that it is more suitable for parametrs estimation than DKE?

Thanks for all of your advices. I appreciate it.

Hi. I personally used DKE for DKI maps estimation. However, its installation and usage relies on MATLAB2012 compiler and it is thus obsolete and I couldn’t install it on my MAC. That is why I started using DESIGNER, which is also easy to implement as you just have to use:
python DESIGNER.py -denoise -degibbs -rician -DKIparams input output

1 Like

Hi ros, thanks for your advice. Meanwhile I download DESIGNER.py and because my data are already preproc by MRtrix3, I decided to use just tensorfitting.m function in matlab. When I use option to detect outliers with IRWLLS, I keep getting warnings for parpool and for Matrix is singular, close to singular or badly scaled. Results may be inaccurate. RCOND = NaN. I am not sure why.
Have you ever tried to run just this one function, not whole script in python? Thanks for your time, MB.

That is expected. You are not masking out background during the tensor estimation. So there are background voxels where you are fitting DKI to no signal, just noise. Of course the fit is ill conditioned in background so this warning is produced. It is generally nothing to worry about and it does not mean that your maps are not accurate.

Dear ros,
now I can see it :smiley: Oh! How stupid question.

Thanks for your answer. I run DESIGNER.py for DKI and DTI params. My DTI outputs looks pretty ok, but in DKI maps I keep seeing some kind of drop outs/holes/dark dots all around the brain. I am not sure, such as I it my first DKI dataset processing. It makes me thinking whether there is something uncommon in my data or it is due to processing (I think second makes no sense since I am not doing any changes in code and everything is installed ok). I will put here my mean kurtosis image.

Sorry for spam here in MRtrix community, but I don’t know whether there is any community for Designer project.

Thanks for all of your help.