Hi,
I'm not entirely sure what is happening here, but it is possible that
some NaNs in the original data might be causing the problem.
You can run it through fslmaths initially to remove NaNs by doing:
fslmaths inputimage -nan outputimage
and then run this through melodic again. Hopefully that will fix it.
If not, let us know.
All the best,
Mark
On 1 Jul 2009, at 18:34, Sunil Narayan wrote:
> Hi, I'm attempting to run MELODIC ICA on some fmri data I pre-
> processed in
> AFNI, I used 3dAFNItoNIFTI to convert the 4D AFNI files to NIFTI and
> ran FSL
> -> MELODIC turning the motion correction off and leaving the
> functional
> brain extraction and highpass filter on, the report seems to load
> the data
> fine, and then gives an empty mask image error after which, all the
> values
> are reported as not a number. Just wondering if there is something
> simple I
> need to do here. I read an early post with a similar error related
> to byte
> swapping, but after examining the data in AFNI and the nii file in
> fsl view
> the values appear to be identical.
> Thanks
> -Sunil
>
> Verbatim Report:
>
> Prestats
>
> /misc/apps/fsl-4-centos5_64/bin/fslmaths
> /misc/radium0/narayans/BreathingVisualTask/s09/Sunilprocessing/
> filt01_MotRVT_RTRCR_pc_st_reg_s09.rest.run1
> prefiltered_func_data
> Total original volumes = 300
>
> /misc/apps/fsl-4-centos5_64/bin/fslroi prefiltered_func_data
> example_func 150 1
>
> /misc/apps/fsl-4-centos5_64/bin/bet prefiltered_func_data
> prefiltered_func_data_bet -F
>
> /misc/apps/fsl-4-centos5_64/bin/fslmaths prefiltered_func_data_bet -
> thrp 10
> -Tmin -bin mask -odt char
>
> /misc/apps/fsl-4-centos5_64/bin/fslmaths prefiltered_func_data_bet -
> mas mask
> prefiltered_func_data_bet
>
> /misc/apps/fsl-4-centos5_64/bin/fslmaths mask -kernel gauss
> 2.12314225053
> -fmean mask_weight -odt float
>
> /misc/apps/fsl-4-centos5_64/bin/fslmaths prefiltered_func_data_bet -
> kernel
> gauss 2.12314225053 -fmean -div mask_weight -mas mask
> filtered_func_data
> -odt float
>
> /misc/apps/fsl-4-centos5_64/bin/fslmaths filtered_func_data -ing 10000
> filtered_func_data -odt float
> ERROR:: Empty mask image
> ERROR:: Empty mask image
> ERROR:: Empty mask image
> <repeats couple hundred times>
>
> /misc/apps/fsl-4-centos5_64/bin/fslmaths filtered_func_data -bptf
> 25.0 -1
> filtered_func_data -odt float
>
> /misc/apps/fsl-4-centos5_64/bin/fslmaths filtered_func_data -Tmean
> mean_func
>
> /bin/rm -rf prefiltered_func_data*
> No convergence after 500 steps
>
>
>
> An exception has been thrown
> Runtime error:- detected by Newmat: process fails to converge
>
> MatrixType = Diag # Rows = 4; # Cols = 4; lower BW = 0; upper BW = 0
> Trace: Evalue(tql2).
>
>
> Melodic Version 3.05
>
> Melodic results will be in filtered_func_data.ica
>
> Create mask ... done
> Reading data file filtered_func_data ... done
> Estimating data smoothness ... done
> Removing mean image ... done
> Excluding voxels with constant value ... done
>
> Normalising by voxel-wise variance ... done
>
> Data size : 300 x 0
>
> Starting PCA ... done
> Start whitening using 1 dimensions ...
> retaining nan percent of the variability
> ... done
>
> Starting ICA estimation using symm
>
> Step no. 1 change : nan
> Step no. 2 change : nan
> <all nan down to step 499>
> Step no. 499 change : nan
>
> Restarting MELODIC using -d 4
>
> Starting PCA ... done
> Start whitening using 4 dimensions ...
> retaining nan percent of the variability
> ... done
>
> Starting ICA estimation using symm
>
> <Registration didn't show any errors...>
>
> Thanks
> -Sunil
>
|