Hi Chris
Are you sure that you have non-zero values in the medial wall? That was basically my problem with a similar error.
Best, Blazej
On 13 Apr 2017, at 17:40, Christopher Markiewicz <[log in to unmask]> wrote:
> Hi all,
>
> I'm wondering if anybody else has run across (or can help me diagnose) this singular matrix issue for surface -- but not volume -- GLMs. My attempts to search the list have been unproductive, and Blazej's thread turned out to be for a different issue.
>
> Thanks again,
> Chris Markiewicz
>
> ________________________________________
> From: FSL - FMRIB's Software Library <[log in to unmask]> on behalf of Christopher Markiewicz <[log in to unmask]>
> Sent: Thursday, April 6, 2017 1:32:44 PM
> To: [log in to unmask]
> Subject: [FSL] Singular matrix when running film_gls --mode=surface but not for equivalent volumetric analysis
>
> Hi all,
>
> I'm working on a preprocessing stream that will produce data that is generally amenable to different analysis streams, including FSL. One component of this is sampling functional volumes to FreeSurfer meshes, and I want to test this component by running a simple first-level GLM with film_gls, so that I can compare the output to a volumetric GLM.
>
> I've previously run the following volumetric analysis, which runs as expected:
>
> film_gls --noest --fcon=run0.fts --rn=results --con=run0.con --in=mni_bold.nii.gz --pd=run0.mat --thr=0.000000
>
> Here `mni_bold.nii.gz` is derived from a `native_bold.nii.gz`. Additionally, we sample `native_bold.nii.gz` to the subject mesh, and transform to the `fsaverage` mesh, producing `fsaverage_bold.L.func.gii`.
>
> film_gls --noest --fcon=run0.fts --rn=results --con=run0.con --in=fsaverage_bold.L.func.gii --pd=run0.mat --thr=0.000000 --in2=smoothwm.fsaverage.L.surf.gii --mode=surface
>
> Abbreviated output follows after the signature, but the upshot is that an exception is thrown due to a singular matrix. It's unclear to me where this issue is arising; if the design matrix were singular, it should produce an error in the volumetric analysis.
>
> Thanks in advance for your help.
>
> --
> Chris Markiewicz
> Center for Reproducible Neuroscience
> Stanford University
>
>
> ----------------------
> Standard output
> ----------------------
> Log directory is: results
> read gifti
> read scalar data that is float32
> ... [x 180]
> paradigm.getDesignMatrix().Nrows()=180
> paradigm.getDesignMatrix().Ncols()=13
> sizeTS=180
> numTS=163842
> Completed
> Prewhitening and Computing PEs...
> Percentage done:
>
> ----------------------
> Standard error
> ----------------------
> ** DA[0] has coordsys with intent NIFTI_INTENT_TIME_SERIES (should be NIFTI_INTENT_POINTSET)
> ...
> ** DA[179] has coordsys with intent NIFTI_INTENT_TIME_SERIES (should be NIFTI_INTENT_POINTSET)
>
>
> An exception has been thrown
> Runtime error:- detected by Newmat: matrix is singular
>
> MatrixType = Crout # Rows = 3; # Cols = 3
> Trace: Crout(lubksb); GeneralSolvI; InvertedMatrix::Evaluate.
|