Dear Ross,
>
> Thanks so much for the suggestion.
>
> I re-ran the commands with --ol_ec=2 as you suggested, and they completed successfully.
>
> I found a very brief explanation of this otherwise-hidden eddy option in an earlier email from you:
>
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=fsl;12467485.1811
>
> It sounds like this flag shouldn't be used routinely, but kept as an alternative if this problem occurs with datasets that have only a small number of volumes with the lowest b-value?
that has definitely been my thinking. It is a tricky question though, and it is possible that that recommendation might change in the future.
>
> I'm surprised that you wrote "not being able to find a single volume in the b=500 shell that doesn’t have at least one outlier slice" - so, ideally, its better to have few outlier slices for this procedure? That seems non-intuitive - or are you expecting a certain level of motion? (At least, it might upset our "professional" subjects who pride themselves on not moving in the scanner!)
What I have seen in some data sets is that certain MB-groups has been consistently a little higher/lower than the other groups. Presumably due to sequence dependent differences in the time between subsequent excitations for the different groups. These differences are very small, and probably of no consequence. But they can be interpreted by eddy as outliers, especially when there are low b-val shells in the context of high b-val shells, rendering the outlier detection very sensitive for those low b-val volumes. So it is not necessarily down to subject movement.
>
> On a somewhat related note - I've also noticed that the cnr_maps file being created by eddy with s2v are ending up with NaNs in the file; while when the same data is processed without s2v correction but otherwise identical parameters, this is not the case.
>
> I get NaNs if I use fslstats on the files, and eddy_quad is logging:
> !!!Warning!!! NaNs detected in the CNR maps!!!
> although it goes on to complete the QC report. The SNR and CNR maps in the QC report look almost identical whether s2v was used or not - so its not immediately obvious where the NaN voxels are, and whether they can be safely ignored.
>
Hmm, I don’t think I have seen that myself. Have you checked for NaN in the images you input to eddy?
Jesper
> Thanks,
>
> Ross
>
> ------------------------------
>
> Date: Mon, 9 Dec 2019 14:22:27 +0000
> From: Jesper Andersson <[log in to unmask]>
> Subject: Re: EDDY::: Unable to find volume with no outliers in shell 0 (was: Re: eddy: msg=ECScanManager::set_slice_to_vol_reference: ref index out of bounds)
>
> Hi again Ross,
>
>
>> I am getting a very similar error message to the last poster (Nikitas); ie:
>>
>> EDDY::: DoVolumeToVolumeRegistration: Unable to find volume with no outliers in shell 0 with b-value=49?
>> EDDY::: Eddy failed with message EDDY::: eddy.cpp::: EDDY::ReplacementManager*
>> EDDY::DoVolumeToVolumeRegistration(const EDDY::EddyCommandLineOptions&, EDDY::ECScanManager&): Exception
>> thrown
>>
>> This is occurring when I run eddy_cuda with slice-to-volume correction; however when I run eddy_cuda without s2v
>> correction, there are no problems.
>>
>> In this dataset, I also have a small number of volumes - 6 - with low-b-value ~ 500. Does this mess with the s2v process? > (No volumes with b ~ 50 - so not sure where it found that number!)
>
> as I suspected this is an issue with not being able to find a single volume in the b=500 shell that doesn’t have at least one outlier slice. It seems that for some reason the final number it the error message is overwritten with a ?.
>
> I can’t re-run your command here because I don’t have the "slspec_ABCD102.txt” file. Can I ask you to please re-run the s2v command with --ol_ec=2 and let me know what you get?
>
> Jesper
>
> ########################################################################
>
> 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
|