NODDI outputs from AMICO quality check

[Qianlan Chen]

Hi all,

I’ve been applying the AMICO on my multi-shell DWI data to get the NODDI maps. I’d like some feedback regarding the quality of my output.

Looks like there are some black voxels on ODI in white matter region, I was wondering if it is due to the location of CST(Corticospinal tract) or improper fitting, as these voxels’ value is very close to 0 or even 0 on ODI? Comparing to ICVF, the corresponding WM region looks fine for us (cf. fig1 and fig2).

Before fitting the DWI data with the NODDI model using AMICO, we did two steps for denoising: 1) Gibbs’ unringing (mrdegibbs in MRtrix3) and 2) Rician noise removal (one configuration setting in AMICO). Based on the output, we further got the modulated ICVF and ODI maps based on tissue weighted means. But still those black voxels remain even on modulated ODI map (cf. fig3 and fig4).

Besides this issue, we would like to ask for your comments regarding the quality of our NODDI output, seems like there’s residual noise in our images (especially in ODI)(cf. fig5).

Here are our multi-shell DWI protocol parameter as below, further parameters in fig6:

TE: 89.20ms

TR: 3230ms

Loading data:

        * DWI signal...

                - dim    = 140 x 140 x 92 x 99

                - pixdim = 1.500 x 1.500 x 1.500

        * Acquisition scheme...

                - 99 samples, 12 shells

                - 7 @ b=0 , 14 @ b=1490.0 , 5 @ b=2990.0 , 10 @ b=1495.0 , 9 @ b=3005.0 , 13 @ b=1500.0 , 8 @ b=2985.0 , 9 @ b=2995.0 , 12 @ b=3000.0 , 8 @ b=1505.0 , 2 @ b=2980.0 , 1 @ b=1485.0 , 1 @ b=3010.0

 

Any suggestions or comments would be highly appreciated.

 

Qianlan

       

 

 

[Hui Zhang]

Hi Qianlan,

The issue you are experiencing is likely caused by having 12 "shells" but you really only have two.  AMICO has an option for you to round off the b-values, such that 1490.0 becomes 1500.

I suggest that you try that.

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.
To view this discussion on the web visit https://groups.google.com/d/msgid/noddi/810d99a1-612f-4f06-a8c4-081314a25b6bn%40googlegroups.com.

[Qianlan Chen]

Hi Gary,

Thank you so much for your suggestions. I just rounded off the b-value by setting "amico.util.fsl2scheme(bvals, bvecs, bStep=100)" in the scheme, here is the attached protocol output from AMICO and the results: 

As shown in the attached picture,  both results looked very similar and the voxel value as well when I checked (upper row the previous, and the lower row is the latest result).  Would you have any other suggestions to try to improve our NODDI maps? 

Thank you so much!

Best, 

Qianlan

-> Loading data:
        * DWI signal...
                - dim    = 140 x 140 x 92 x 100

                - pixdim = 1.500 x 1.500 x 1.500
        * Acquisition scheme...

                - 100 samples, 2 shells
                - 7 @ b=0 , 47 @ b=1500.0 , 46 @ b=3000.0
        * Binary mask...

                - dim    = 140 x 140 x 92

                - pixdim = 1.500 x 1.500 x 1.500

                - voxels = 419625

[Hui Zhang]

Hi Qianqian,

Can you provide the difference map?

By the way, from the AMICO command line output messages, it appears that the data has gone from 99 volumes to 100.  Just in case you  weren’t aware.

Gary

To view this discussion on the web visit https://groups.google.com/d/msgid/noddi/645bde93-a842-4ffc-af61-d992c5777f7fn%40googlegroups.com.