Epoching of continuous eeg source localization

Discussions
,
1

Hi everyone,

I have an EEG dataset including 53-channel EEG data recorded simultaneously while users performed muscular tasks, such as pushing a sensor with their middle finger, for 20 trials. The data in each task were recorded continuously. Each trial included both rest and task conditions: in the rest condition, the subject did not perform any task for 3 seconds, and in the task condition, the subject performed a specific muscular task for 5 seconds.

I preprocessed my data using EEGLAB, and now I have my preprocessed EEG data. Currently, I am working with the Brainstorm toolbox to extract source-level data from the sensor data.

I have read the tutorial and other related documents provided, but I have one question. Since my dataset isn't like ERP data that requires epoching and averaging trials, I would like to know if it is correct to perform source localization analysis on all the data samples without applying any epoching?

For more detail, my sampling frequency is 1200 Hz. In each trial, I have 3fs samples related to the rest condition and 5fs samples related to the task condition, so in each trial, I have 9600 samples. In total, I have 20*9600 samples for one subject.

Is it technically correct if I load all these samples and then calculate the source time series, or must I do it for a small time window separately (one trial)?

2

In principle, most source localization algorithms are independent of time. The solutions depend solely on the EEG values at a certain instant, meaning that you could do source localization of the complete time series.

However, this is not advisable, since the signal-to-noise ratio of the event can be greatly improved by averaging the events. If you use source localization on the averaged event, the results are expected to be better.

I would thus advise you to cut the full signal into epochs containing the task and not the task. Then average those to an average for the task and the non-task. Finally, source localize those averages.