Hi Natàlia, the 2018 and 2009 should give nearly identical results, particularly for the mixed head model. The 2016 update by me did not include mixed head models, which I have fixed now. I would welcome your testing and confirming or noting any noticeable changes.
The 2016 update also changed the definition of the noise regularization parameter to be a fraction of the largest eigenvalue, but this was causing confusion with users who thought it should be the fraction of the average eigenvalue of the noise covariance matrix. Both the 2009 and the 2018 now use noise regularization parameter as the average of the eigenvalues, which is the same as the average of the diagonal values.
Because of these differences, we opted for now to leave all three versions up, so that users can adapt and test the new 2018 interface, which will ultimately become the default.
- John, One of the Brainstorm Principal Investigators