Dear Helmut,
many thanks for your time and testing with this issue. Indeed, this is highly relevant in my field, since extra-cerebral structures of rat and mice have a high uptake of the PET / SPET tracers and therefore can have a big influence in the results if the masking is not properly applied.
What I am wondering now (and I will try to test later) is if the results of applying directly a mask to the image are different if the masked values are replaced by zeros or if replaced by NaN.
Regards,
David Vállez
> El 17/11/2014, a las 21:27, Helmut Nebl <[log in to unmask]> escribió:
>
> To conclude, it seems the global calculation / mean
> 1) is based on all non-NaN voxels of the images, thus it does matter whether images have been masked during preprocessing, even if these masked voxels are not part of the analysis
> 2) does ignore NaN voxels
> 3) works on the images and does ignore explicit/implicit masks and thresholds given during model estimation
>
> At least when using these scripts, which seem to reflect SPM default settings. Hope this is correct now. In any case, it might be useful to clarify these issues in the manual, as it might have a huge impact on the analysis depending on settings.
>
> Best,
>
> Helmut
|