Brain cells communicate with electricity. This electricity can be measured at the scalp with electrode sensors and recorded to create a time-series dataset.
Participants in EEG studies usually wear nets containing multiple EEG electrode sensors. These nets are designed so that electrodes make contact with the scalp at specific locations, enabling this to be standardised across participants (example of a simple montage below).
EEG studies typically ask participants to complete a series of short simple tasks in one sitting while wearing an electrode net with a standard montage.
The open dataset we are using in this hackathon consists of EEG data collected from 60 participants (of which we have included 48 subjects) wearing EEG nets each containing 64 electrode sensors (of which we have included 61).
In this dataset, participants completed three separate recording sessions in total (this is typically performed to enable longitudinal analysis, or to investigate reliability).
In each recording session, participants completed five tasks:
Dataset reference: Yulin Wang and Wei Duan and Lihong Ding and Debo Dong and Xu Lei (2021). A test-retest resting and cognitive state EEG dataset. OpenNeuro. [Dataset] doi: doi:10.18112/openneuro.ds003685.v1.0.5. Dataset licence: CC0
EEG data is “noisy”. Brain signals are very small (typically 10 µV to 100 µV at the scalp) and contain many larger unwanted signals picked up by the EEG electrodes. Unwanted signals usually include electrical interference from electricity mains, and also biological electrical signals generated by the eyes and muscles. The image below shows what noise would look like in the signal (highlighted by yellow boxes).