Dear Phillip,
your problem might arise from signal drop out in the affected regions
leading to lower numerical values, which are in turn eliminated by the
masking threshold (fmri_spec.mthresh = 0.8). You can try to lower this
threshold to, e.g., 0.2, or set it to -Inf (no adaptive masking) since
you are using a brain mask anyway. However, be careful with statistical
results in these regions.
Best,
Manfred
Am 20.11.2017 um 00:31 schrieb Phil W.:
> Dear SPM’ers,
>
> I’m having a problem at very early stages of my analysis pipeline and I can’t seem to find an explanation for it. I’m trying to process partial brain volumes in SPM12 in MATLAB2016a(9.0). To put it simply, at some point during the level one analysis, large parts of midbrain and some pre-frontal cortex is being cut out of the individual subject’s mask. The attached pictures show the problem visually.
>
> Myself and others in the lab have asked colleagues what the problem could be, and the general answer people come up is that there is a problem in normalization. However, as you can see from picture 1 (Original EPI_ Norm EPI.jpg) the normalization process doesn’t seem to effect the signal we’re getting from the EPI – that is, it doesn’t seem to be cutting anything out. Also, the normalized EPI seems to come out well aligned to the normalized T1 (picture 2; Norm T1_Norm EPI.jpg). However, I’ve attached a code snipet from our normalization batch code, just in case.
>
> We think the problem lies in the level 1 analysis. As you can see in the subject’s mask produced by level 1 analysis, large parts of the midbrain and prefrontal cortex are chopped out despite clear signal in these regions from EPI. The mask we use at the beginning of level 1 is the standard SPM12 mask (mask_ICV.nii). We’ve also attached a snipet from our level 1 batch.
>
> Any help would be very much appreciated!
>
> Best,
> Phillip Witkowski
|