Dear Carlos,
looks like this correction is only used if you change your modality from "fmri" to "PET". Anyway, you can try to increase the absolute mask threshold up to 0.2-0.25 and check that no important effects are cut.
If you have used CAT12 for your preprocessing I would _always_ use the provided T1 i(Template_T1.nii or Template_T1_masked.nii) in CAT12 for overlay. The single subject brain in SPM!2 is never a good choice (even if you have used SPM12 for preprocessing), because it's just one single brain and it deviates from the space that is used in CAT12 (MNI152NLin2009cAsym).
Best,
Christian
On Mon, 22 Aug 2022 18:44:37 +0000, Carlos Murillo Ezcurra <[log in to unmask]> wrote:
>Dear SPM-CAT12 developers/users,
>
>I am new user of SPM12-VBM/CAT12 and I wanted to ask for advice about an issue with a possible false positive cluster (at uncorrected level) due to low variance outside the brain (attached).
>I have used an absolute threshold mask of 0.1 (fwhm 6). Design: independent t-test (63 vs 32 controls). I have read the paper where this issue is addressed and solution is proposed:
>
>Ridgway, G. R., Litvak, V., Flandin, G., Friston, K. J., & Penny, W. D. (2012). The problem of low variance voxels in statistical parametric mapping; a new hat avoids a �haircut�. Neuroimage, 59(3), 2131-2141.
>
>And I found that the proposed solution is implemented in the spm function for the Estimation of a General Linear Model: spm_spm.m
>
>Line 656
>
>%==========================================================================
>
>%- R e s M S M O D I F I C A T I O N
>
>%==========================================================================
>
>
>
>%-Modify ResMS (a form of shrinkage) to avoid problems of very low variance
>
>try
>
> if ~strcmpi(spm_get_defaults('modality'),'fmri')
>
> ResMS = spm_data_read(VResMS);
>
> ResMS = ResMS + 1e-3 * max(ResMS(isfinite(ResMS)));
>
> VResMS = spm_data_write(VResMS, ResMS);
>
> clear ResMS
>
> end
>
>end
>
>My question: Is this modified residual mean squares variance estimate automatically implemented when we call the 2nd level analysis in the SPM GUI? If not, how I could implement this correction?
>If this modification is actually implemented, what would be the approach to take (a 0.2 mask do not solve the issue)?
>The problematic cluster, in particular, does not survive once I correct for multiple comparisons but I am still curious about the implications.
>
>Thanks so much in advance for your help!
>Regards
>
>Carlos Murillo
>PhD student
>
>[signature_1588322910]
>
>[A picture containing text, outdoor, sign Description automatically generated] [A picture containing text, clipart Description automatically generated]
>
>Ghent University, Faculty of Medicine and Health Sciences
>Department of Rehabilitation Sciences
>SPINE Research Unit
>Campus UZ Gent, Corneel Heymanslaan 10, B3, 9000 Ghent, Belgium
>Secretary phone number +32 9 332 26 32
>
>
|