Back to the main page.

Bug 1044 - MNE does not work smoothly in the context of singletrial/rawtrial

Reported 2011-10-17 09:55:00 +0200
Modified 2012-01-03 11:29:29 +0100
Product: FieldTrip
Component: inverse
Version: unspecified
Hardware: PC
Operating System: Mac OS
Importance: P1 normal
Assigned to: Jan-Mathijs Schoffelen
Depends on:
See also:

Jan-Mathijs Schoffelen - 2011-10-17 09:55:30 +0200

it seems that single trial noise covariances are supported -> with rawtrial. As such this does not really make sense.

Jan-Mathijs Schoffelen - 2011-12-21 16:10:00 +0100

I suggest to change ft_sourceanalysis such that it works the same way for the pcc/dics/lcmv and for the mne. This entails the support for keeping filters in mne, and allowing for cfg.singletrial to be activated for mne. In this context, pre-computed filters (based on the average noise covariance) need to be in the input. This entails also support for using precomputed filters in mne

Jan-Mathijs Schoffelen - 2011-12-21 16:12:51 +0100

Sorry, where I said cfg.singletrial in the previous post, I intended to say 'rawtrial'

Jan-Mathijs Schoffelen - 2011-12-21 16:24:29 +0100

implemented with revision 5066. The logic for mne is now the same as with the beamformers: first compute the spatial filters using the average, and then reconstruct the single trials using the pre-computed filters in cfg.grid.filter = source.avg.filter, and cfg.rawtrial = 'yes';

Arjen Stolk - 2011-12-21 16:35:00 +0100

Great! This is now the same as the method described for DICS: (In reply to comment #3)