Question about CHARMED method

[alestella]

Hi developers, thanks so much for sharing qMRLab with the community!
I have a question on the CHARMED method.
Referring to the guide at https://qmrlab.readthedocs.io/en/master/charmed_batch.html#36, the method assumes that the diffusion gradients are applied perpendicularly to the neuronal fibers.
How can I then use it with my diffusion dataset including, let's say, 40+ diffusion directions?
Also, I don't understand where to insert the information of the q-values or b-values.

Thanks so much for your help.

[agahkarakuzu]

Hi @alestella,

First, you'll need to convert your bvecs and bvals as in:

bvecfile = 'dwi.bvec'; % FSL bval text file (output of dcm2nii or dicm2nii)
bvalfile = 'dwi.bval'; % FSL bvec text file (output of dcm2nii or dicm2nii)
Gmax = 40; % maximal gradient strength of the MRI system used in mT/m
scd_schemefile_FSLconvert(bvalfile, bvecfile, Gmax, 'scheme.txt')

This script above will create scheme.txt, that you can use with the Load button at the Protocol/DiffusionData panel.

Regarding the orthogonality of gradient directions with the neurites, it is an assumption of the CHARMED model, nothing to control for (if anyone could, the acquisition would be pointless) while choosing diffusion directions. That is not to say there are not any recommended practices to sample directions for a CHARMED (multishell) acquisition:

This is from the original article (https://doi.org/10.1016/j.neuroimage.2005.03.042), and it ensures that for each b value, the diffusion gradient directions were uniformly and symmetrically distributed over a sphere (see this link for reference).

How can I then use it with my diffusion dataset including, let's say, 40+ diffusion directions?

As you can see in the table above, for using CHARMED, you need a large range of b-values and gradient directions. Actually, there is a whole article dedicated to the optimization of gradient sampling scheme for CHARMED acquisitions.

Bottomline, first you need to ensure that the data you have in hand is suitable for this model. Ideally, the question should be "how do I need to collect this data so that it satisfies the requirements of CHARMED?".

I hope these are useful, good luck!

[alestella]

Thanks so much for the thorough and quick reply, @agahkarakuzu!
Thank you also for the precious suggestions! I think I need to study more in depth the appropriate literature before diving in into the analysis - it is the correct way to approach scientific problems - my bad!
For practical reasons I did not explain in much details, but I do have an increasing number of directions with different b-shells.
I will double-check the quality of the dataset though, and the compatibility with the requirements you mentioned.

Thanks so much again, and have a nice day and rest of the week!

[tanguyduval]

Hi @alestella and Agah,

I hope you are doing well :)
In this implementation your gradients have to be orthogonal to the fiber direction. Which mean that gradient direction must be in a plane.
It is adapted to structures where you know fiber direction at acquisition.

If you have 3d shells for your diffusion gradients, you should use NODDI for instance, which is also a multi compartment model (based on charmed).

Concerning the appellations:
CHARMED has been used with 3D shells, but this not the original implementation as implemented here.
When you fit the diameter of the axons, charmed is called AxCaliber.

Hope that help,
Tanguy

[agahkarakuzu]

Thank you @tanguyduval !

Great to see that you are still keeping an eye on here :)

[alestella]

Thanks so much @tanguyduval for the clarifications!

The parameter I am interested in is in particular the fraction of restricted diffusion (Fr, in the CHARMED model). I suppose I shall look at the NODDI neurite density index, instead. I admit I was looking on the web for some quick tool to analyze my data and see whether the Fr would tell me something - that is why I originally thought of using qMRlab.

Thank you also for clarify about the AxCaliber and the 3D-shell implementation of CHARMED!