Synchronise multiple signal

[Edouard2laire]

Hello,

This process aims at aligning multiple signals acquired with different devices but sharing synchronizing events (typically sent to each recording device at the time of the acquisition). This situation can arise easily when recording multimodal data such as EEG - iEEG, EEG - fMRI, fNIRS - fMRI, or EEG - fMRI.

Having the signal synchronised makes making easier to perform multimodal analysis, notably allowing the simultaneous exploration of the different signals. For example, here is a figure I recently presented at a conference, showing how we can understand better what is happening during a seizure by looking at the EEG, fNIRS, and pulse oximeter

What's left to do:

• Fix the code to save continuous files as output
• Improve file history

Future work

Here is a list of future things we might consider implementing but that is not required for this PR

• Allowing the simultaneous visualization of the different signals from Brainstorm (opening multiple continous viewer at the same time. #656)
• Explicitly linking the different synchronized files. This could notably be used so that if an event is added to a file, then the same event would be added to all the events the file is synchronized with.
• Add the option to either crop the longer signal keeping only the common signal (current implementation) or pad the shorter signal with either 0 or NaN

Let me know if you want me to send you an example dataset. I can generate a minimum brainstorm database with 2 files to synchronize.

Regards,
Edouard

[rcassani]

Let me know if you want me to send you an example dataset. I can generate a minimum brainstorm database with 2 files to synchronize.

@Edouard2laire, yes, please send me a minimal example dataset

[Edouard2laire]

Hi.
DId you had time to review this PR and #656 @rcassani ?

Thanks a lot,
Edouard

[rcassani]

@Edouard2laire, apologies for the delay. there was a misunderstanding, I thought the PR still was on development. I will review it ASAP

[rcassani]

@Edouard2laire, the main change in the process is that files are loaded one by one in memory.

Minor changes:

• It now removes the completely out-of-range events, and fixes the partially out extended events.
• List of synch files in History

[rcassani]

The alignment could be improved by using a Gaussian around the occurrences.

[Edouard2laire]

Thanks a lot for all the changes!
I'll test it this weekend and let you know if everything works.

Edouard

[Edouard2laire]

Hello.

I did few test and it seems to be working. Thanks a lot,

Edouard

[rcassani]

@Edouard2laire, great!

[Edouard2laire]

(Clicked wrong button — you can ignore that)

[rcassani]

For the open tasks:

• The feature of multiple raw time-series plots can be handled in another PR.
• Signals are now chopped to be the same length, not sure if the alternative of zero-padding would be useful
• Synching other events (after synching signals) for the moment can be done with Transfer events

[Edouard2laire]

For the open tasks:

• The feature of multiple raw time-series plots can be handled in another PR.

Do you want me to create the PR ? I don't really have any idea on where to start for that implementation.

• Signals are now chopped to be the same length, not sure if the alternative of zero-padding would be useful

yes, I agree that it's not really useful. Could be implemented if someone asks. This can be the case if you still want to keep all the signals and don't have the constraint of having all the recordings all the time.

• Synching other events (after synching signals) for the moment can be done with Transfer events

yes, that's true

[rcassani]

Do you want me to create the PR ? I don't really have any idea on where to start for that implementation.

I'll start the PR, I need to check some implementation of the raw-file readers