Filling the holes is no good idea at all. If you work on the masks after model estimation and fill the holes, then you treat the subjects as if they had a beta of 0 in this particular voxel. But this does not have to be the case at all, just your signal is very low.
You should rather think about why you have signal dropout in some subjects only. Using the same slice orientation should result in rather similar dropouts in general. There are some methods papers dealing with regions which are prone to susceptibilitiy artefacts (amygdala, OFC) and how to reduce dropouts. Maybe a fieldmap can help as well.
|