Hi Christian,
I haven't figured out how to make melodic work with the bigger dataset I
have, but the ICA (not PCA) of the concatenated IC maps looks impressive:
the maps are anatomically well-defined and match very well with the ones you
reported in your paper.
I had tried also another approach, clustering the spatial correlation matrix
between pairs of individual IC maps, but the ICA^2 (!) approach seem to do a
much better job.
Thanks for all your suggestions, and if you happen to find out how to avoid
the bad_alloc error please let me know (I would like to see whether those
RSNs replicate or change in my treatment sample...)
very best
giuseppe
On Wed, 10 May 2006 23:29:24 +0100, Giuseppe Pagnoni <[log in to unmask]> wrote:
>Hi Christian,
>
>thank you very much for your reply. I tried your suggestion, and while it
>works on the smaller of two datasets (a *.nii.gz file of ~ 26 MB), it throws
>an alloc error when using a larger dataset (a *.nii.gz file of ~ 37 MB):
>
> Melodic Version ln(12) beta
>
>Melodic results will be in run2.IC.cat.MEDT.ica
>
>Reading data file run2.IC.cat.MEDT ... done
>Create mask ... done
>Excluding voxels with constant value
>terminate called after throwing an instance of 'std::bad_alloc'
> what(): St9bad_alloc
>Abort
>
>Is this a known memory limitation of melodic? I tried it both on a FreeBSD
>machine and a Fedora 5 machine, with the same result.
>
>thanks again for your help
>
> giuseppe
>=========================================================================
|