>1. Where do you apply the explicit mask, when defining the model (i.e. “Specify 1st-level” module in SPM)
here
>or when looking at the results (i.e., “Results” module in SPM)?
in results, the mask hide the results only
>If it is the former, then SPM should screen those voxels that exceed the threshold (i.e., 0.4) within your explicit mask, right?
no ; there is a confusion.
masking procedure in the 1st level analysis is done to eliminate voxels of no interest from the analysis, and so, decrease the multiple comparison problem in the following statistical step.
the implicit mask (functional) is not an image but a procedure thresholding the functional images to keep only voxels above 40%of the mean intensity value (or 80% by default).
the explicit mask does’nt depend on the functional images but only on a mask image that you give to the model. You can choose “inclusive or exclusive” option, meaning that you want the analysis inside or outside the mask image.
I choose both masks.
>2. Considering that functional images are generally smoothed, some researchers suggest smoothing the explicit mask as well. Do you agree with a such opinion?
No. but for my anatomical mask, I’d prefer to have a low threshold of the proba of grey matter (often 50%).
If you are doing group analysis, an alternative is to functionally mask the individual data (1st level) and anatomical restrict the analysis at the group level. With the fmriprep processing, the anatomical normalization is very good, so the averaged grey matter of the group is still lovely and so you can threshold this mean image to have a group anatomical mask for the group level, which represent the best mean probability to have GM in each voxel….
|