clarification of what event manipulation versus epoching based on events mean
TO DO :
for the app-epoch:
Add in the report.html how many epochs were created (for each of the events selected and in total). This is important because nb of events can be different from the nb of epochs creating (if there is an issue with the time window selected for the epoch creation for example)
move tmin and tmax into the main parameter definition window
add optional parameters in advanced options: baseline subtraction (with time window)
add optional parameters in advanced options: reject, flat, reject_tmin, reject_tmax
create the event-related app(s): app-event:
create at least the visualize and extract one.
Examine the possibilities of event-manipulation (creating event, combining event, etc) that MNE offers. Contact Laurent HUGUEVILLE who has worked on this a lot for the GOGAIT pipeline.
investigate what remains to be done with regards to maxfilter / head position correction:
in some cases, we have continuous head pos recording. Then head movement correction is to be done at the level of the single run.
(also some low-pass filtering is needed because of the high-frequency artifacts introduced by the small currents injected into the coils to localize them (aka. to localize the head)).in other cases, we don’t have continuous head pos recording, but we have one head pos per run, and often we have several runs per subjects, with different head positions in each run, and we want to be able to combine/pool together these runs. Hence we need to realign all runs on a common, reference position in order to then pool them together. For this, we need i) to compute a common (e.g. mean) head position for all the runs ii) realign the runs on this common head position. This is the purpose of some apps that Aurore did. It seems that the MNE BIDS-pipeline has also some options to take care of this. This is to be investigated and implement.
Ultimately, the two types of head position correction (within runs, when there is continuous head pos recording, and across runs when we want to align several runs on a common, fixed head position) have to be implemented.
continue working on the decoding app. We will test it on a dataset of GEORGE Nathalie with the aim of a scientific paper.