Hi,

I'm using 'mri_glmfit' with surface inputs. However, from subject to subject, some vertices have zero-values. One solution is to use the 'prune' flag to keep only non-zeros vertices across subjects. However, this solution seems too conservative as too many vertices are discarded, producing gruyère-like significance maps.

Is it possible to adopt a softer version of prune, i.e., only do the inference on non-zeros frames, without including frames that have null signal? In other words, this solution would consist in having different degrees of freedom throughout the surface.

thanks,

julien