Hello Alessandro,
It is difficult to say what is going on here - we would probably need an upload of your 1st and 2nd level FEAT directories to examine.
Kind Regards
Matthew
> Dear FSL experts
>
> We have been using FEAT to map the effect of psychoactive drugs in the mouse/rat brain using CBV-based fMRI - a procedure that normally works very well, and that we have been using successfully for years.
> The procedure is based on a first pass analysis where we fit a regressor of interest to the treatment groups (typically drug and "placebo") followed by a 2nd level analysis where we run a simple inter-group comparison.
>
> We have however now a dataset where the effect of a certain drug after 2nd level analysis (drug versus vehicle) is not coming out in the maps as usual, these being suspiciously "grainy" and having an "unsmoothed" appearance. This is in sharp contrast with what we see on first-level maps, where the effect of the drug is robust and widespread (and smoothed), and with ROI based analysis which confirms a robust and widespread effect of the compound.
>
> Basically, the effect on the maps is where it should be, only it does not appear to be smoothed, with a grainy appearance which we normally do not see and that we find unappealing. This is the case even after using very high 1st level smoothing (e.g 5 or 10 voxels). We tried OLS, FLAME different threshold levels, voxel correction vs cluster correction, different levels of brain/background threshold all to no avail. The original 4D datasets look "normal" and devoid of artefacts to us.
>
> Is there anyone out there who could give us a clue of what could be going on, and some hints of how to troubleshoot this?
>
> Thanks in advance
> Alessandro
>
|