search

finnlennartsson / kpum_noddi

repo for processing of DKI/NODDI data for preterm project at KPUM
0 stars 0 forks source link

DKI estimation #5

Closed finnlennartsson closed 1 year ago

finnlennartsson commented 1 year ago

Find a way to do DKI estimations.

Can be put within dmri_preprocess.sh where the DTI fit is made. https://github.com/finnlennartsson/kpum_noddi/blob/fcb31888c15ebe662da77da914419c8e4464b464/scripts/dmri_preprocess.sh#L368-L369

finnlennartsson commented 1 year ago

MRtrix do not support DKI estimations yet (it is in Dev Branch) - see https://community.mrtrix.org/t/kurtosis/6413?u=finn

Perhaps it can be solved by applying FSL - https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FDT/UserGuide#DTIFIT

NOTE that we should use the WLS fit (this is done per default in MRtrix dwi2tensor).

finnlennartsson commented 1 year ago

Actually, best to only use one shell for estimation of tensor elements. See below from FSL mailing list:

Hi Yes you will get FA/MD values predicted by the kurtosis model. But one issue with FA in particular is that it may be biased due to the assumption that you have one single kurtosis parameter controlling non-gaussian diffusion along all directions. So I would use single shell at least for FA calculations. Cheers Saad

From: FSL - FMRIB's Software Library [FSL@JISCMAIL.AC.UK](mailto:FSL@JISCMAIL.AC.UK) on behalf of Jun Yeon Won [won25@UMD.EDU](mailto:won25@UMD.EDU) Date: Wednesday, 25 January 2023 at 01:57 To: FSL@JISCMAIL.AC.UK [FSL@JISCMAIL.AC.UK](mailto:FSL@JISCMAIL.AC.UK) Subject: Re: [FSL] Multishell Data dtifit Question

Hi Saad, Thank you for the explanation. Could you clarify one more thing? If I use the —kurt option for the multishell data, the FA and other diffusion parameters are calculated based on the kurtosis model already? Best, Jun

finnlennartsson commented 1 year ago

One can also check out https://pydesigner.readthedocs.io/en/latest/index.html for DTI/DKI calculations. But seems to want to do the whole pipeline, which is tricky in our case.

finnlennartsson commented 1 year ago

Or Diffusion_Kurtosis_Estimator - https://www.nitrc.org/projects/dke/ I have downloaded, but it requires Matlab

finnlennartsson commented 1 year ago

Or DIPY. Here are instructions: https://dipy.org/documentation/1.6.0./examples_built/reconst_msdki/#example-reconst-msdki

finnlennartsson commented 1 year ago

Have implemented this using instructions on DIPY's dipy_fit_dki

This is implemented in https://github.com/finnlennartsson/kpum_noddi/blob/616f9457162233945eae09099a77a319da5216a0/shell/dmri_dtidki.sh#L140-L156