Ex Vivo imaging

[farshid sepehrband]

Hi,

I would like to use NODDI for ex vivo imaging of human brain. I was wondering if you have any suggestion regarding the acquisition protocol. I presume due to low water diffusivity, higher b-value would be better. Is there a literature on NODDI for ex vivo imaging?

Regards,

Farshid

[Hui Zhang]

Hi Farshid,

Thanks for asking.  Yes, I have indeed worked with a number of collaborators on such application.  I suggest that you have a look at the following poster which illustrates the work led by Michiel Kleinnijenhuis.

http://f1000.com/posters/browse/summary/1093706

Gary

--
You received this message because you are subscribed to the Google Groups "NODDI" group.
To unsubscribe from this group and stop receiving emails from it, send an email to noddi+un...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[farshid sepehrband]

Hi Gary,

Thanks for your help. 

I was wondering which model I use for fitting NODDI to ex-vivo. I noticed that there is fourth component (stationary water) for ex-vivo cases. I would appreciate if you let me know what should be used instead of 'WatsonSHStickTortIsoV_B0'.

Regards,

Farshid

[farshid sepehrband]

I think I found the answer. Is it WatsonSHStickTortIsoVIsoDot_B0 ? 

[Hui Zhang]

That is it.

Gary

For more options, visit https://groups.google.com/d/optout.

[farshid sepehrband]

Hi Gary

Thanks for your helps. 

I have fitted NODDI to our ex-vivo data. One of the outputs is called irfrac.nii. What is it? if it is related to volume fraction of one of the compartments, how can I formulate them considering that we have four compartment now. For example, for three compartments case, we have, (1-Fiso)Fic, which is neurite volume fraction (Please correct me if I'm wrong).

Regards,

Farshid

That is it.

To unsubscribe from this group and stop receiving emails from it, send an email to noddi+unsubscribe@googlegroups.com.

[Hui Zhang]

Hi Farshid,

The model includes the compartments in a hierarchical way, for example with the original model, at the 1st level: you have signal contributions from the tissue and the CSF.  At the 2nd level, the signal from tissue is subdivided into the part from the intra-neurite and the one from the extra-neurite.

So when I say neurite density, I am referring to the fraction of neurite within tissue, i.e. at the 2nd level, not how you have understood, which gives the total fraction of neurite within a voxel.  The reason that I prefer the first way is that it allows one to exclude the effect of partial volume from CSF.

Now onto the ex vivo NODDI model, we have now a isotropically restricted compartment, also known as the DOT compartment.  Please refer to the reference Alexander 2010 in the poster.  irfrac refers to the fraction of this compartment.  It is implemented in a slightly unexpected way perhaps, as the 0th level, i.e., above the 1st  and 2nd levels.

Gary

For more options, visit https://groups.google.com/d/optout.

[Beth Hutchinson]

Hi Gary,
I wanted to follow up on this string to make sure I understand the ex-vivo model correctly.  The poster you referenced is helpful and so is looking through the code, but I couldn't find an equation that explicitly states these relationships (please forgive me if I have overlooked it though).

Is this correct?

1=V(irfrac) + V(iso) + V(hindered)*[V(ic)+V(ec)]

where V(irfrac) is the restricted compartment volume ratio, V(iso) is CSF volume ratio and V(hindered) is the hindered compartment volume ratio, which is divided into intracellular (ic) and extracellular (ec) so that their volume ratios sum to 1, i.e. V(ic)+V(ec)=1.

Thanks for your help.
Beth

[Hui Zhang]

Hi Beth,

Your equation is not quite right.  We implemented our multi-compartment in a completely hierarchical fashion:

At the top level, we split the total signal into two

1) isotropic fraction (Virfrac) + non isotropic fraction (1 - Virfrac)

Next, we split the isotropic fraction into two

2) free diffusion (Viso) + non free diffusion (1 - Viso)

Finally, we split the non free diffusion into two

3) intra-cellular diffusion (Vic) + extra-cellular diffusion (1 - Vic)

extra-cellular diffusion is modelled as the hindered diffusion (with a diffusion tensor); there isn't a separate V(hindered).

So theoretically, Virfrac, Viso, and Vic can be equal to 1 all at the same time, or all equal to 0.

Gary

[Beth Hutchinson]

Thank you, so would it be this:

1 = (1-Virfrac)+Virfrac*[Viso+(1-Viso)*[Vic+(1-Vic)]]

or could it be:

1 = Virfrac+(1-Virfrac)*[Viso+(1-Viso)*[Vic+(1-Vic)]]

[Hui Zhang]

1 = Virfrac+(1-Virfrac)*[Viso+(1-Viso)*[Vic+(1-Vic)]]

This is the correct interpretation.

Gary

[Beth Hutchinson]

Thank you, very helpful!

[wjju...@gmail.com]

Hi Gary

I tried to use NODDI for ex-vivo imaging but I'm not sure that I used it properly.

I used 9.4 T Agilent MRI to took image of ex-vivo white matter and Gmax was 288mT/m. I took 3 shells of DTI sequence such as 36 directions at b=3000, 72 directions at b=5000, 121 directions at b=8000. small del=10ms, large del=16ms.

resolution was 0.2*0.2*2(mm^3), TE was 28.2ms, used AMICO for NODDI fitting

As you can see on my zip file, dti_fa map is pretty reliable but I don't know why the FIT_ICVF map looks so unclear compared to your Ex-vivo NODDI fitting.

Also, one of weird thing is the value of white matter at FIT_ICVF. Its range is between 0.18~0.34 and it seems like way too low compared to In-vivo WM FIT_ICVF.

All of these things makes me doubt that I used NODDI for Ex-vivo image properly.

Could you give any advice to me?  I would attach the data to here.

https://drive.google.com/file/d/0B6UtCvntxS79MTNYM0Y3S1B1VDQ/view?usp=sharing

Thanks!

[Hui Zhang]

Hello,

I am not sure AMICO at the moment has the built-in support for ex vivo NODDI fitting.  This might explain the odd results you got.

The NODDI toolbox does support ex vivo NODDI fitting.  But you will need to use a different model called "WatsonSHStickTortIsoVIsoDot_B0".  Let me know if this works for you.

Gary

--

[wjju...@gmail.com]

Hi Gary,

I followed your advice so used the model called 'WatsonSHStickTortIsoVIsoDot_B0'.(also, changed the maximum G in 'FSL2Protocol' to 288mT/m that I used at 9.4T MRI experiment)

but as you can see, fitting doesn't seem to work properly.

Here is my data which I uploaded on my Google Drive.
https://drive.google.com/file/d/0B6UtCvntxS79dHF4c1FFNzVEV3c/view?usp=sharing

If there's anything wrong, plz let me know.

Thanks!!

WooJin