Dear Genevieve,
<insert here a paragraph about the use of SPM8 in 2021> ;-)
You are doing the right thing by first checking the first level contrast
images. If they look OK to you (brain coverage, intensities) then I
wouldn't investigate option B (for the record, "masking threshold" can
be lowered but shouldn't really be set to -Inf unless one has very good
reasons to do so).
Is this a "standard" fMRI analysis? Have you done any unusual step
during the preprocessing? If nothing comes to mind, the easiest might be
to share your contrast images with me so that I can have a look.
Best regards,
Guillaume.
On 23/03/2021 22:54, Genevieve Allaire-Duquette wrote:
> Dear SPM experts,
>
> I'm analyzing a data set using SPM8 and Matlab R2016b on Mac.
>
> Preprocessing and first-level analysis are running smoothly, but L2 analysis generates the following error:
>
> "Failed 'Model estimation'
> Error using spm_spm (line 831)
> Please check your data: There are no significant voxels.
> In file "/Applications/spm8/spm_spm.m" (v4515), function "spm_spm" at line 831.
> In file "/Applications/spm8/config/spm_run_fmri_est.m" (v4403), function "spm_run_fmri_est" at line 33.
>
> No executable modules, but still unresolved dependencies or incomplete module inputs."
>
> From what I could gather on the forum, the error could be caused by two main issues:
>
> A. the majority of the voxels have been EXCLUDED when running group spec, suggesting some contrast images are EMPTY. -> I checked all first-level contrast images used to run L2 analyses using CheckReg, and I don't see any evidence that a significant number of voxels have been excluded.
>
> B. the effect of implicit masking. SPM uses the mean BOLD signal at each voxel to determine "in-brain" vs. "out-of-brain" voxels so this may be causing some issues in my analyses. -> recommendation is to try setting the "masking threshold" value to "-inf" when defining first-level analysis in SPM (its default value is "0.8") and see whether that helps.
>
>
> Would you recommend I investigate option B? If so, how can I change the threshold to "-inf".
>
> Any other tips?
>
>
> Thanks a lot for your help.
>
> Genevieve Allaire-Duquette
>
--
Guillaume Flandin, PhD
Wellcome Centre for Human Neuroimaging
UCL Queen Square Institute of Neurology
London WC1N 3BG
|