Import Raw Data
Importing raw data means opening your EEG recording in analysis software so you can start working with it. EEG recordings come in several file formats, including BioSemi .bdf, European Data Format .edf, BrainVision files, and EEGLAB .set files. Some recordings use a group of files: BrainVision typically has a .vhdr header, an .eeg data file, and a .vmrk event-marker file, while an EEGLAB .set file may have a separate .fdt data file. Keep these companion files together when moving or importing a recording. MNE’s guide to EEG file formats explains these common examples.
For a continuous recording whose EEG channels share the same sampling rate, a helpful way to picture the imported data is as a big table, or matrix. Each row represents one EEG channel, each column represents a successive time sample, and each cell contains the voltage measured by that channel at that moment. Reading across a row follows one channel through time; reading down a column compares the channels at one moment. The files themselves can store this information in different ways, but a channel-by-time matrix is a common representation used for analysis, as illustrated in MNE’s explanation of continuous data.
When we plot one channel as an EEG trace, time is on the horizontal x-axis and voltage is on the vertical y-axis, usually displayed in microvolts (µV). The voltage readings are the values inside the matrix, rather than its row labels. A microvolt is one millionth of a volt. The stored numbers are not always already in microvolts: formats such as BDF store digital values with scaling information, and software such as MNE uses volts internally before scaling EEG plots to microvolts.
This matrix can get large quickly. The sampling rate tells us how many measurements are recorded per second for each channel. For example, 64 EEG channels recorded at 1,000 Hz for one minute give us 60,000 time samples per channel: a 64 × 60,000 matrix containing 3.84 million voltage readings. More channels, a higher sampling rate, or a longer recording all mean more values to load and process. Importing also brings in available recording information, such as channel names, sampling rate, and event markers, which helps us interpret those numbers and later identify the parts of the recording we want to analyze.