neurokit2 0.1.3

The Python Toolbox for Neurophysiological Signal Processing.

The Python Toolbox for Neurophysiological Signal Processing

This package is the continuation of NeuroKit 1. It’s a user-friendly package providing easy access to advanced biosignal processing routines. Researchers and clinicians without extensive knowledge of programming or biomedical signal processing can analyze physiological data with only two lines of code.

Quick Example

import neurokit2 as nk

# Download example data
data = nk.data("bio_eventrelated_100hz")

# Preprocess the data (filter, find peaks, etc.)
processed_data, info = nk.bio_process(ecg=data["ECG"], rsp=data["RSP"], eda=data["EDA"], sampling_rate=100)

# Compute relevant features
results = nk.bio_analyze(processed_data, sampling_rate=100)

And boom 💥 your analysis is done 😎

Installation

You can install NeuroKit2 from PyPI

pip install neurokit2

or conda-forge

conda install -c conda-forge neurokit2

If you’re not sure what to do, read our installation guide.

Contributing

NeuroKit2 is the most welcoming project with a large community of contributors with all levels of programming expertise. But the package is still far from being perfect! Thus, if you have some ideas for improvement, new features, or just want to learn Python and do something useful at the same time, do not hesitate and check out the following guides:

• Understanding NeuroKit

• Contributing guide

• Ideas for first contributions

Documentation

Click on the links above and check out our tutorials:

General

• Get familiar with Python in 10 minutes

• Recording good quality signals

• What software for physiological signal processing

• Install Python and NeuroKit

• Included datasets

• Additional Resources

Examples

• Simulate Artificial Physiological Signals

• Customize your Processing Pipeline

• Event-related Analysis

• Interval-related Analysis

• Analyze Electrodermal Activity (EDA)

• Analyze Respiratory Rate Variability (RRV)

• Extract and Visualize Individual Heartbeats

• Locate P, Q, S and T waves in ECG

• Complexity Analysis of Physiological Signals

• Analyze Electrooculography EOG data

• Fit a function to a signal

You can try out these examples directly in your browser.

Don’t know which tutorial is suited for your case? Follow this flowchart:

Citation

The NeuroKit2 paper can be found here 🎉 Additionally, you can get the reference directly from Python by running:

nk.cite()

You can cite NeuroKit2 as follows:

- Makowski, D., Pham, T., Lau, Z. J., Brammer, J. C., Lespinasse, F., Pham, H.,
Schölzel, C., & Chen, S. A. (2021). NeuroKit2: A Python toolbox for neurophysiological signal processing.
Behavior Research Methods. https://doi.org/10.3758/s13428-020-01516-y

Full bibtex reference:

@article{Makowski2021neurokit,
    author={Makowski, Dominique and Pham, Tam and Lau, Zen J. and Brammer, Jan C. and Lespinasse, Fran{\c{c}}ois and Pham, Hung and Sch{\"o}lzel, Christopher and Chen, S. H. Annabel},
    title={NeuroKit2: A Python toolbox for neurophysiological signal processing},
    journal={Behavior Research Methods},
    year={2021},
    month={Feb},
    day={02},
    issn={1554-3528},
    doi={10.3758/s13428-020-01516-y},
    url={https://doi.org/10.3758/s13428-020-01516-y}
}

Let us know if you used NeuroKit2 in a publication! Open a new discussion (select the NK in publications category) and link the paper. The community would be happy to know about how you used it and learn about your research. We could also feature it once we have a section on the website for papers that used the software.

Physiological Data Preprocessing

Simulate physiological signals

import numpy as np
import pandas as pd
import neurokit2 as nk

# Generate synthetic signals
ecg = nk.ecg_simulate(duration=10, heart_rate=70)
ppg = nk.ppg_simulate(duration=10, heart_rate=70)
rsp = nk.rsp_simulate(duration=10, respiratory_rate=15)
eda = nk.eda_simulate(duration=10, scr_number=3)
emg = nk.emg_simulate(duration=10, burst_number=2)

# Visualise biosignals
data = pd.DataFrame({"ECG": ecg,
                     "PPG": ppg,
                     "RSP": rsp,
                     "EDA": eda,
                     "EMG": emg})
nk.signal_plot(data, subplots=True)

Electrodermal Activity (EDA/GSR)

# Generate 10 seconds of EDA signal (recorded at 250 samples / second) with 2 SCR peaks
eda = nk.eda_simulate(duration=10, sampling_rate=250, scr_number=2, drift=0.01)

# Process it
signals, info = nk.eda_process(eda, sampling_rate=250)

# Visualise the processing
nk.eda_plot(signals, sampling_rate=250)

Cardiac activity (ECG)

# Generate 15 seconds of ECG signal (recorded at 250 samples / second)
ecg = nk.ecg_simulate(duration=15, sampling_rate=250, heart_rate=70)

# Process it
signals, info = nk.ecg_process(ecg, sampling_rate=250)

# Visualise the processing
nk.ecg_plot(signals, sampling_rate=250)

Respiration (RSP)

# Generate one minute of respiratory (RSP) signal (recorded at 250 samples / second)
rsp = nk.rsp_simulate(duration=60, sampling_rate=250, respiratory_rate=15)

# Process it
signals, info = nk.rsp_process(rsp, sampling_rate=250)

# Visualise the processing
nk.rsp_plot(signals, sampling_rate=250)

Electromyography (EMG)

# Generate 10 seconds of EMG signal (recorded at 250 samples / second)
emg = nk.emg_simulate(duration=10, sampling_rate=250, burst_number=3)

# Process it
signal, info = nk.emg_process(emg, sampling_rate=250)

# Visualise the processing
nk.emg_plot(signals, sampling_rate=250)

Photoplethysmography (PPG/BVP)

# Generate 15 seconds of PPG signal (recorded at 250 samples / second)
ppg = nk.ppg_simulate(duration=15, sampling_rate=250, heart_rate=70)

# Process it
signals, info = nk.ppg_process(ppg, sampling_rate=250)

# Visualize the processing
nk.ppg_plot(signals, sampling_rate=250)

Electrooculography (EOG)

# Import EOG data
eog_signal = nk.data("eog_100hz")

# Process it
signals, info = nk.eog_process(eog_signal, sampling_rate=100)

# Plot
plot = nk.eog_plot(signals, info, sampling_rate=100)

Electrogastrography (EGG)

Consider helping us develop it!

Physiological Data Analysis

The analysis of physiological data usually comes in two types, event-related or interval-related.

Event-related

This type of analysis refers to physiological changes immediately occurring in response to an event. For instance, physiological changes following the presentation of a stimulus (e.g., an emotional stimulus) indicated by the dotted lines in the figure above. In this situation the analysis is epoch-based. An epoch is a short chunk of the physiological signal (usually < 10 seconds), that is locked to a specific stimulus and hence the physiological signals of interest are time-segmented accordingly. This is represented by the orange boxes in the figure above. In this case, using bio_analyze() will compute features like rate changes, peak characteristics and phase characteristics.

• Event-related example

Interval-related

This type of analysis refers to the physiological characteristics and features that occur over longer periods of time (from a few seconds to days of activity). Typical use cases are either periods of resting-state, in which the activity is recorded for several minutes while the participant is at rest, or during different conditions in which there is no specific time-locked event (e.g., watching movies, listening to music, engaging in physical activity, etc.). For instance, this type of analysis is used when people want to compare the physiological activity under different intensities of physical exercise, different types of movies, or different intensities of stress. To compare event-related and interval-related analysis, we can refer to the example figure above. For example, a participant might be watching a 20s-long short film where particular stimuli of interest in the movie appears at certain time points (marked by the dotted lines). While event-related analysis pertains to the segments of signals within the orange boxes (to understand the physiological changes pertaining to the appearance of stimuli), interval-related analysis can be applied on the entire 20s duration to investigate how physiology fluctuates in general. In this case, using bio_analyze() will compute features such as rate characteristics (in particular, variability metrics) and peak characteristics.

• Interval-related example

Heart Rate Variability (HRV)

If you are looking for:

• a comprehensive review of the most up-to-date HRV indices

• a discussion of their significance in psychological research and practices

• a step-by-step guide for HRV analysis using NeuroKit2

the Heart Rate Variability in Psychology: A Review of HRV Indices and an Analysis Tutorial paper is a good place to start.

You can cite the paper as follows:

- Pham, T., Lau, Z. J., Chen, S. H. A., & Makowski, D. (2021).
Heart Rate Variability in Psychology: A Review of HRV Indices and an Analysis Tutorial.
Sensors, 21(12), 3998. https://doi:10.3390/s21123998

Full bibtex reference:

@article{Pham_2021,
    author={Pham, Tam and Lau, Zen Juen and Chen, S. H. Annabel and Makowski, Dominique},
    title={Heart Rate Variability in Psychology: A Review of HRV Indices and an Analysis Tutorial},
    volume={21},
    ISSN={1424-8220},
    url={http://dx.doi.org/10.3390/s21123998},
    DOI={10.3390/s21123998},
    number={12},
    journal={Sensors},
    publisher={MDPI AG},
    year={2021},
    month={Jun},
    pages={3998}
}

• Compute HRV indices

  • Time domain: RMSSD, MeanNN, SDNN, SDSD, CVNN etc.

  • Frequency domain: Spectral power density in various frequency bands (Ultra low/ULF, Very low/VLF, Low/LF, High/HF, Very high/VHF), Ratio of LF to HF power, Normalized LF (LFn) and HF (HFn), Log transformed HF (LnHF).

  • Nonlinear domain: Spread of RR intervals (SD1, SD2, ratio between SD2 to SD1), Cardiac Sympathetic Index (CSI), Cardial Vagal Index (CVI), Modified CSI, Sample Entropy (SampEn).

# Download data
data = nk.data("bio_resting_8min_100hz")

# Find peaks
peaks, info = nk.ecg_peaks(data["ECG"], sampling_rate=100)

# Compute HRV indices
nk.hrv(peaks, sampling_rate=100, show=True)
>>>    HRV_RMSSD  HRV_MeanNN   HRV_SDNN  ...   HRV_CVI  HRV_CSI_Modified  HRV_SampEn
>>> 0  69.697983  696.395349  62.135891  ...  4.829101        592.095372    1.259931

Miscellaneous

ECG Delineation

• Delineate the QRS complex of an electrocardiac signal (ECG) including P-peaks, T-peaks, as well as their onsets and offsets.

# Download data
ecg_signal = nk.data(dataset="ecg_3000hz")['ECG']

# Extract R-peaks locations
_, rpeaks = nk.ecg_peaks(ecg_signal, sampling_rate=3000)

# Delineate
signal, waves = nk.ecg_delineate(ecg_signal, rpeaks, sampling_rate=3000, method="dwt", show=True, show_type='all')

Signal Processing

• Signal processing functionalities

  • Filtering: Using different methods.

  • Detrending: Remove the baseline drift or trend.

  • Distorting: Add noise and artifacts.

# Generate original signal
original = nk.signal_simulate(duration=6, frequency=1)

# Distort the signal (add noise, linear trend, artifacts etc.)
distorted = nk.signal_distort(original,
                              noise_amplitude=0.1,
                              noise_frequency=[5, 10, 20],
                              powerline_amplitude=0.05,
                              artifacts_amplitude=0.3,
                              artifacts_number=3,
                              linear_drift=0.5)

# Clean (filter and detrend)
cleaned = nk.signal_detrend(distorted)
cleaned = nk.signal_filter(cleaned, lowcut=0.5, highcut=1.5)

# Compare the 3 signals
plot = nk.signal_plot([original, distorted, cleaned])

Complexity (Entropy, Fractal Dimensions, …)

• Optimize complexity parameters (delay tau, dimension m, tolerance r)

# Generate signal
signal = nk.signal_simulate(frequency=[1, 3], noise=0.01, sampling_rate=100)

# Find optimal time delay, embedding dimension and r
parameters = nk.complexity_optimize(signal, show=True)

• Compute complexity features

  • Entropy: Sample Entropy (SampEn), Approximate Entropy (ApEn), Fuzzy Entropy (FuzzEn), Multiscale Entropy (MSE), Shannon Entropy (ShEn)

  • Fractal dimensions: Correlation Dimension D2, …

  • Detrended Fluctuation Analysis

nk.entropy_sample(signal)
nk.entropy_approximate(signal)

Signal Decomposition

# Create complex signal
signal = nk.signal_simulate(duration=10, frequency=1)  # High freq
signal += 3 * nk.signal_simulate(duration=10, frequency=3)  # Higher freq
signal += 3 * np.linspace(0, 2, len(signal))  # Add baseline and linear trend
signal += 2 * nk.signal_simulate(duration=10, frequency=0.1, noise=0)  # Non-linear trend
signal += np.random.normal(0, 0.02, len(signal))  # Add noise

# Decompose signal using Empirical Mode Decomposition (EMD)
components = nk.signal_decompose(signal, method='emd')
nk.signal_plot(components)  # Visualize components

# Recompose merging correlated components
recomposed = nk.signal_recompose(components, threshold=0.99)
nk.signal_plot(recomposed)  # Visualize components

Signal Power Spectrum Density (PSD)

# Generate complex signal
signal = nk.signal_simulate(duration=20, frequency=[0.5, 5, 10, 15], amplitude=[2, 1.5, 0.5, 0.3], noise=0.025)

# Get the PSD using different methods
welch = nk.signal_psd(signal, method="welch", min_frequency=1, max_frequency=20, show=True)
multitaper = nk.signal_psd(signal, method="multitapers", max_frequency=20, show=True)
lomb = nk.signal_psd(signal, method="lomb", min_frequency=1, max_frequency=20, show=True)
burg = nk.signal_psd(signal, method="burg", min_frequency=1, max_frequency=20, order=10, show=True)

Statistics

• Highest Density Interval (HDI)

x = np.random.normal(loc=0, scale=1, size=100000)

ci_min, ci_max = nk.hdi(x, ci=0.95, show=True)

Popularity

NeuroKit2 is one of the most welcoming package for new contributors and users, as well as the fastest growing package. So stop hesitating and hop onboard 🤗

Notes

The authors do not provide any warranty. If this software causes your keyboard to blow up, your brain to liquefy, your toilet to clog or a zombie plague to break loose, the authors CANNOT IN ANY WAY be held responsible.

News

0.1.3

Breaking Changes

• None

New Features

• Add internal function for detecting missing data points and forward filling missing values in nk.*_clean() functions

• Add computation of standard deviation in eventrelated() functions for ECG_Rate_SD, EMG_Amplitude_SD, EOG_Rate_SD, PPG_Rate_SD, RSP_Rate_SD, RSP_Amplitude_SD

• Add labelling for interval related features if a dictionary of dataframes is passed

• Retrun Q peaks and S Peaks information for wavelet-based methods in nk.ecg_delineate()

Fixes

• Fix epochs columns with dtype: object generated by nk.epochs_create()

• Bug fix ecg_findpeaks_rodrigues for array out of bounds bug

0.1.2

New Features

• Additional features for nk.rsp_intervalrelated(): average inspiratory and expiratory durations, inspiratory-to-expiratory (I/E) time ratio

• Add multiscale entropy measures (MSE, CMSE, RCMSE) and fractal methods (Detrended Fluctuation Analysis, Correlation Dimension) into nk.hrv_nonlinear()

• Allow for data resampling in nk.read_bitalino()

• Add bio_resting_8min_200hz into database for reading with nk.data()

• Reading of url links in nk.data()

• Allow for nk.hrv() to compute RSA indices if respiratory data is present

• All hrv functions to automatically detect correct sampling rate if tuple or dict is passed as input

• Add support for PPG analysis: nk.ppg_eventrelated(), nk.ppg_intervalrelated(), nk.ppg_analyze()

• Add Zhao et al. (2018) method for nk.ecg_quality()

• Add tests for epochs module

• Add sub-epoch option for ECG and RSP event-related analysis:

  • users can create a smaller sub-epoch within the event-related epoch

  • the rate-related features of ECG and RSP signals are calculated over the sub-epoch

  • the remaining features are calculated over the original epoch, not the sub-epoch

Fixes

• Fix propagation of values in nk.signal_formatpeaks() for formatting SCR column outputs generated by eda_peaks()

• Fix docstrings of nk.rsp_phase(), from “RSP_Inspiration” to “RSP_Phase”

• Update signal_filter() method for rsp_clean(): to use sos form, instead of ba form of butterworth (similar to eda_clean())

0.1.1

New Features

• Use duration from nk.events_find() as epochs_end in nk.epochs_create()

• Allow customized subsets of epoch lengths in nk.bio_analyze() with window_lengths argument

• Add nk.find_outliers() to identify outliers (abnormal values)

• Add utility function - nk.check_type() to return appropriate boolean values of input (integer, list, ndarray, pandas dataframe or pandas series)

• (experimental) Add error bars in the summary plot method to illustrate standard error of each bin

Fixes

• Fix type of value in nk.signal_formatpeaks() to ensure slice assignment is done on the same type

0.0.1 (2019-10-29)

• First release on PyPI.