fICVF is always 1?

[Manish Amin]

hello,

I am performing a study on rat spinal cords and have collected multishell data (5 shells ranging from 2900-13000 bval with varying mixing times).  When i process the data however, I am getting some extremely odd values.  I am seeing fiso values ranging from .1 - .9 (possible but I cant believe it would ever get that high), and the ficvf value is always 1. considering these are both fractions, they obviously don't add up properly, which means there must be something I did wrong. Has anyone ever had this problem before? I believe I am following the directions properly from the website.  the ODI images show some nice GW/WM contrast, and so does the fiso images, so why is the ficvf values always equal to 1?

-Manish

[Manish Amin]

I made some images to show what I am referring to.  these are NODDI outputs of a spinal cord that we took.

[Hui Zhang]

Hi Manish,

Yes, it is strange to have FICVF to be uniformly 1.  Are you scanning your sample ex vivo?  If so, have you used the corresponding model 'WatsonSHStickTortIsoVIsoDot_B0' instead, as described previously here?

https://groups.google.com/d/msg/noddi/22XwFxMi_i4/IMGq-JWrVz0J

I should also note that your assumption about the fractions adding to 1 is NOT consistent with how the model has been constructed, and described, in my paper.  In fact, FISO and FICVF are two independent variables, each of which can vary from 0 to 1.  Take a look at Figure 2 of the following paper:

https://www.sciencedirect.com/science/article/pii/S1053811916000616

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+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Manish Amin]

ah that explanation makes sense. I apologize, I should have been more careful when reading the literature.  

These samples are ex vivo. I did not previously use this model, but I did this time, and these are the results I got. They still don't seem to make sense.  Also, the fiso and ODI values now seem to not make sense.

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

[Hui Zhang]

Can you produce the time series view for a few voxels in GM/WM, as explained in my tutorial here?

https://www.nitrc.org/frs/download.php/10193/Explore_DWI_Data_with_FSLview.pdf

As well as attaching your bval and bvec files?

Gary

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

[Manish Amin]

Hi Gary,

Sure here are 3 WM voxels and 3 GM voxels' time series, as well as the attached bvals and bvecs. 

[Hui Zhang]

Hi Manish,

Thanks for sending these.  They confirmed my suspicion.  You mentioned in one of your earlier emails that you used a range of mixing times.  That suggested to me that your DWI measurements at different mixing times can not be trivially combined.  Pay attention to the b=0 measurements in particular.  You can clearly see that they are dramatically different between the first two and the last two.  So you need to first normalise your DWI data of each mixing time before combining.  To normalise your DWI data, you need to first compute the average of the b=0 volumes, then, for each voxel, divide each DWI measurement (including b=0) with the average b=0 value.

Gary

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

[Manish Amin]

Ah I cannot believe I didn't realize that to begin with, that was dumb of me.  Thank you for catching my error!

[Manish Amin]

Is there an easy way to normalize the diffusion data, such as in FSL or something similar? I can definitely write code to do it but I figured I'd ask first.

[Hui Zhang]

I suspect you can implement the steps I suggested in FSL w/ fslmaths.

Gary

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

[Manish Amin]

Yup I was able to figure it out and the resulting images look a lot better! Thanks for the help!

[Manish Amin]

For anyone that is interested, this normalization was easy to do using the fslroi function to extract the b0 image, and the fslmaths -div option to normalize.

[Hui Zhang]

Good to know it works out.

Gary

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

[Hui Zhang]

If you are up for writing this up as a tutorial, like the one I wrote about DWI, happy to upload it to NITRC for other's benefit.

Gary 

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

[Manish Amin]

Hi Gary,

I apologize I didn't get notifications from this channel, and had no idea you had asked me to do that.  I will write up a tutorial next week and email it to you.

[Manish Amin]

Hi Gary,

I thought i uploaded the tutorial i had written, but I cannot seem to find it on this thread.  Someone else had just asked me about the same issue that we have seen, I was just wondering, did I send you the tutorial i had written? if not, do you still want me to send that?

[Sean Ma]

HI Manish,

Would you be kind enough to upload your tutorial somewhere online (Dropbox, Box) and provide a link to share?

I think it will benefit a lot of followers for NODDI. Thanks!

Sean

[Hui Zhang]

Hi Manish,

I don't believe I have received a copy of your tutorial.

Gary

[Md. Mamun Al-Amin]

Hi Manish,

Would you be able to send the tutorial?

Thanks in advance.

Kind regards

Mamun