Hi Clara,
thanks for the additional information, this is helpful!
If you are calling the scripts with
[lhRgs,lhRgMeans] = lme_mass_RgGrow(lhsphere, lhRe, lhTh0, lhcortex, 2, 95);
then 'lhcortex' will contain the analysis mask, i.e. indices for those vertices for which the analysis will be performed.
'lhcortex' should already be a subset of the vertices specified by 'lhsphere', i.e. excluding some portions of the medial wall.
If you have zero-valued columns in your data matrix, one thing worth trying is to remove the corresponding indices from 'lhcortex', i.e. further restrict the analysis mask. This should be done before the initial covariance estimation, i.e. prior to calling lme_mass_fit_init or lme_mass_fit_EMinit. In addition, it is of course worthwhile to investigate the cause of the zero-columns.
Best,
Kersten