Getting Started

EEGProc is a vectorized library for preprocessing EEG (electroencephalogram) data, extracting features from it, and evaluating models on it subject by subject. It is built for researchers and developers working in neuroscience, biomedical engineering, and machine learning.

Installation

Install from PyPI (Python 3.10 or newer):

pip install eegproc                    # preprocessing, features, data conversion, plotting
pip install "eegproc[deep-learning]"   # adds cross-validation, models, counterfactuals

or, for the latest development version:

pip install git+https://github.com/EEGo-UNC/EEGProc.git

Dependencies

The base install relies on:

  • NumPy, Pandas, SciPy – numerical processing

  • PyWavelets – wavelet features

  • PyEMD (EMD-signal) – empirical mode decomposition

  • Matplotlib – plotting utilities

The deep-learning extra adds TensorFlow, scikit-learn, and cloudpickle.

Quick Start

  1. Import and load your EEG data:

import pandas as pd
from eegproc import bandpass_filter, FREQUENCY_BANDS

df = pd.read_csv("my_eeg_data.csv")
fs = 128  # Hz

  1. Filter into frequency bands:

from eegproc import bandpass_filter, FREQUENCY_BANDS

clean = bandpass_filter(df, fs, bands=FREQUENCY_BANDS)
# -> columns named {channel}_{band}, e.g. AF3_alpha

  1. Extract features:

from eegproc import psd_bandpowers, shannons_entropy, hjorth_params

psd = psd_bandpowers(clean, fs, bands=FREQUENCY_BANDS)   # {channel}_{band}
entropy_df = shannons_entropy(psd)                        # {channel}_entropy
hjorth_df = hjorth_params(clean, fs)                      # {channel}_{band}_activity, ...

shannons_entropy consumes the PSD table, not the raw signal, and returns one value per channel describing how evenly that channel’s energy is spread across bands. The other entropy featurizers follow the same shape: wavelet_entropy consumes wavelet_band_energy, and imf_entropy consumes imf_band_energy.

  1. Visualize results:

from eegproc.plotting import plot_eeg_features

plot_eeg_features(entropy_df, title="Shannon Entropy per Channel", seconds=4.0)

Documentation Structure