Dear Niklas,
>
> Dear Ladies and gentlemen,
nice, I haven’t seen that one before.
> when i was testing my EDDY pipelines i saw that the output is not consistent over several runs.
> So on the exact same input data i get different outputs. This is a problem because we are currently trying to look into repeatabilty of our DWI schemes with subjects coming in for multiple sessions. So we want at least the same results within one session.
>
> Is this a known issue or is it expected (i guess more iterations will minimize differences)?
> It would be more convenient though to have the exact same output because the automatic tests are hash based.
It is an expected behaviour. In order to save time it doesn’t use all voxels in the brain for estimating the hyperparameters of the Gaussian Process. It will select a number (given by --nvoxhp) of random voxels from within the --mask and use that. On different runs that random selection will be different.
The difference in results should be very small though, small enough to not matter for pretty much any application.
There is a way of forcing it to select the same voxels, which is to add --initrand to the command line.
Jesper
>
> This is the eddy cmd that i am running:
> eddy_openmp --verbose --cnr_maps --fep --flm=quadratic --ff=10.0 --acqp=encfile.txt --bvals=dwi_rl.bval --bvecs=dwi_rl.bvec --imain=dwi_rl.nii.gz --index=eddy_index.txt --mask=brain_mask.nii.gz --topup=topup_base --data_is_shelled --resamp=jac --niter=5 --nvoxhp=1000 --out=eddy_corrected --repol --residuals
>
> the verbose output starts like this:
> Reading images
> Filling empty planes
> Performing volume-to-volume registration
> Running Register
> Loading prediction maker
> Evaluating prediction maker model
> Calculating parameter updates
> Iter: 0, Total mss = 16.3184
> Loading prediction maker
> Evaluating prediction maker model
> Calculating parameter updates
> Iter: 1, Total mss = 15.7414 ######## in the second run this value is 15.7429
>
> when i take the difference between the eddy_corrected images and open it with miview i get:
> data = 0.108162 +/- 0.0113977 (min=-1446.6, max=1374.26, stdev=44.097)
>
> If you want the data that i used please tell me where to upload.
> Kind regards,
> Niklas Förster
>
> ########################################################################
>
> To unsubscribe from the FSL list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
|