Computing Source Estimates for Single Trials
Data at the single-trial level are obviously noisier than for evoked responses. You should make sure that your data are appropriately filtered and pre-processed, and bad channels are marked. It is recommended to use empty-room data for the computation of the covariance matrix, and to compute an inverse operator with this new covariance matrix.
At the single-trial level, MNE allows you to extract source estimates for ROIs (or labels). Because the inverse operators are specific to invidivual subjects, these ROIs must be defined in the space of the individual MRIs (not the average).
Morphing of ROIs from the average to the individual space can be done using
mne_morph_labels (options e.g. --from --to --labeldir)
Extracting source estimates for ROIs is done by
mne_compute_raw_inverse (options e.g. --in --inv --labeldir --out)
The output will be a continuous time series for each ROI.
In order to epoch your data with regard to events/triggers in your data, use
mne_epochs2mat (options e.g. --raw --mat --events --event --tmin --tmax --lowpass)
This will produce a description file and an epoch file.
From here on, you are on your own...