Skip to content

Loading EEG Data

AIME LOC supports all major EEG file formats via the MNE-Python backend, plus raw NumPy arrays and CSV files from consumer devices.

Supported Formats

Format Extension Devices / Systems MNE Reader
EEGLAB .set / .fdt Any (exported) read_raw_eeglab()
EDF / EDF+ .edf Clinical, Muse (export) read_raw_edf()
BrainVision .vhdr / .vmrk / .eeg BrainProducts read_raw_brainvision()
BioSemi BDF .bdf BioSemi ActiveTwo read_raw_bdf()
EGI MFF .mff EGI / Magstim read_raw_egi()
CSV .csv Consumer devices numpy.loadtxt()
NumPy array ndarray Any / custom Direct construction

Format is auto-detected from file extension.

Loading from Files

from aime_loc import LOC
from aime_loc.eeg import EEG

loc = LOC()
eeg = EEG(loc)

# Auto-detected format
recording = eeg.load("subject01.set")       # EEGLAB
recording = eeg.load("subject01.edf")       # EDF
recording = eeg.load("subject01.vhdr")      # BrainVision
recording = eeg.load("subject01.bdf")       # BDF
recording = eeg.load("subject01.mff")       # EGI

# Path objects work too
from pathlib import Path
recording = eeg.load(Path("data") / "sub-01" / "eeg" / "nback.set")

Inspecting a Recording

print(recording)
# EEGRecording(eeglab, 64ch, 500Hz, 300.0s, raw)

print(f"Channels: {recording.n_channels}")
print(f"Sample rate: {recording.sfreq} Hz")
print(f"Duration: {recording.duration:.1f}s")
print(f"Channel names: {recording.channel_names[:5]}...")

Loading from NumPy Arrays

For custom pipelines or data from non-standard sources:

import numpy as np

# Shape: (n_channels, n_samples)
data = np.random.randn(32, 128000)

recording = eeg.load(data, sfreq=256)

Required parameter

sfreq is required when loading from NumPy arrays. The SDK cannot infer the sampling rate from raw data.

You can optionally provide channel names:

recording = eeg.load(
    data,
    sfreq=256,
    ch_names=["Fp1", "Fp2", "F3", "F4", "C3", "C4", ...],
)

Loading from CSV

Consumer EEG devices often export CSV files:

# Basic CSV (channels as columns)
recording = eeg.load("session.csv", sfreq=256)

# With device preset (sets channel names + expected sample rate)
recording = eeg.load("meditation.csv", device="muse", sfreq=256)

The CSV loader automatically detects whether channels are rows or columns and transposes if needed.

Device Presets

Built-in presets for popular consumer EEG devices:

Device Preset Name Channels Default Sfreq
Muse 2 / Muse S "muse" 4 (TP9, AF7, AF8, TP10) 256 Hz
OpenBCI Cyton "openbci_cyton" 8 250 Hz
Emotiv EPOC X "emotiv_epoc" 14 256 Hz
Neurosity Crown "neurosity" 8 256 Hz
g.tec Unicorn "gtec_unicorn" 8 250 Hz 
# Device preset auto-sets channel names and expected sample rate
recording = eeg.load("data.csv", device="muse")

# Override sample rate if needed
recording = eeg.load("data.csv", device="muse", sfreq=256)

See Consumer Devices for detailed per-device guides.

Working with MNE

The EEGRecording wraps an MNE Raw object. You can access it directly:

# Get the MNE Raw object
raw = recording.to_mne()

# Use any MNE function
print(raw.info)
raw.plot()                    # MNE interactive plot
raw.compute_psd().plot()      # MNE PSD plot

# After custom MNE processing, return to AIME
recording = eeg.from_mne(raw)

Passing MNE Keyword Arguments

Extra keyword arguments are passed through to the MNE reader:

# Example: load specific channels
recording = eeg.load("subject01.set", preload=True)

Next Steps