I don't suppose you're seeing NaNs specifically in matlab after using
ft_read_cifti?  I don't know of anything else that is expected to add NaNs.

Could you do "wb_command -cifti-reduce <your dconn> MEAN mean.dscalar.nii"
and then do -file-information on that?  If that shows NaNs, I suggest also
running it on the inputs to -cifti-merge-dense to check for NaNs there.

Tim


On Tue, Aug 25, 2015 at 6:32 AM, Matthew George Liptrot <
[email protected]> wrote:

> Hi Stam,
>
> Yep, workbench opens them fine. If I load in the dconn file, a surface and
> the greyordinates.dscalar file, then I can see the connectivity maps as I
> click different parts of the surface.
> Is there a better way to check if the values are reasonable? E.g. Would
> the histograms report if NaNs are present?
>
> Good to hear about the FSL release! I guess though that this is a
> wb_command bug (or a bug in my use of it! :-)
>
> Cheers,
>
> M@
>
> On 24/8/15 14:40 , "Stamatios Sotiropoulos" <
> [email protected]> wrote:
>
> Hi Matthew
>
> No, they are not meant to be there. Can you open and display the dconn
> file in workbench?
>
> FYI, there will be a new FSL release sometime this or the next week. It
> will have some bug fixes in probtrackx2 and some new features. It may be
> the case that this problem disappears with the new version.
>
> Cheers
> Stam
>
>
>
> On 24 Aug 2015, at 08:24, Matthew George Liptrot <[email protected]>
> wrote:
>
> Hi all,
>
> We’ve been using the HCP DWI data to generate dense connectomes in
> ‘matrix3’ format. Following previous advice, we are doing this in multiple
> stages:
>
>    1. Generate 3 separate targets (left surface, right surface,
>    subcortical voxels)
>    2. For each, run probtrackx2 (seed = white matter voxels, target as
>    above) multiple times using the ‘-rseed’ option (1000 streams per voxel as
>    a robustness / computational trade-off).
>    3. The multiple probtrackx2 runs are combined with FSL’s fdt_matrix_merge.
>    This gives us 3 FSL  .dot files.
>    4. Convert each of the three .dot files to .dconn using <wb_command
>    -probtrackx-dot-convert>
>    5. Merge the three .dconn files using <wb_command  -cifti-merge-dense>
>    to obtain the final 92K x 92K dconn matrix.
>
> The problem is that we get both NaN’s and zeros in the final, merged dconn
> matrix. Presumably we should not have any NaN’s, as the numbers should just
> represent count statistics?
>
> We looked into the outputs at different stages, and it seems that the
> NaN’s only appear after Step 5.
>
> So our question is: are the NaN’s meant to be there, and if so, what do
> they represent? If not, is this a bug or a mistake on our part?
>
> Many thanks for any help!
>
> M@
>
>
> <http://about.me/matthewliptrot>
> --
> *Matthew George Liptrot*
>
> <http://about.me/matthewliptrot>
> *Department of Computer Science*
> *University of Copenhagen*
> &
> *Section for Cognitive Systems*
> *Department of Applied Mathematics and Computer Science*
> *Technical University of Denmark*
>
> http://about.me/matthewliptrot
>
> _______________________________________________
> HCP-Users mailing list
> [email protected]
> http://lists.humanconnectome.org/mailman/listinfo/hcp-users
>

_______________________________________________
HCP-Users mailing list
[email protected]
http://lists.humanconnectome.org/mailman/listinfo/hcp-users

Reply via email to