NEURAL-py-EEG 0.1.4

A Neonatal EEG Pre-Processing and Feature Set in python

NEURAL_py: A Neonatal EEG Feature Set in python

Python code that replicates the MATLAB version of the NEURAL feature set. The code is used to generate a set of quantitative features from multichannel EEG recordings. Features include amplitude measures, spectral measures, and basic connectivity measures (across hemisphere's only). Also, for preterm EEG (assuming gestational age < 32 weeks), will generate features from bursts annotations (e.g. maximum inter-burst interval). Burst annotations require a separate package, also available on github.

Full details of the methods can be found in: JM O’Toole and GB Boylan (2017). NEURAL: quantitative features for newborn EEG using Matlab. ArXiv e-prints, arXiv:1704.05694 available at https://arxiv.org/abs/1704.05694.

---

Requirements | Use | Quantitative features | Files | Test computer setup | Licence | References | Contact

Requirements:

Python 3.7 or newer with the following packages installed

Package	Version installed
mne	0.18.2
pandas	0.25.1
numpy	1.17.1
matplotlib	3.1.1
scipy	1.3.1
pyEDFlib	0.1.14

Use

Set path (path_to_NEURAL_py) to the folder location of NEURAL_py. Then the following will load the main functions:

  import sys
  sys.path.append(path_to_NEURAL_py)
  import utils,  preprocessing_EEG, generate_all_features

As an example, generate simulated EEG and calculate relative spectral power, standard deviation of range-EEG, and brain symmetry index:

  # generate EEG-like data (coloured Gaussian noise)
  data_st = utils.gen_test_EEGdata(5*60,64,1)

  # define feature set (or can define in neural_parameters.m):
  feature_set = ['spectral_relative_power','rEEG_SD', 'connectivity_BSI']
  
  # estimate features:
  feats_per_epochs, feat_pd_names, feat_st, feats_median_ch, feats_median_all = generate_all_features.generate_all_features(data_st, feat_set=feature_set)

  print('Features extracted')
  print(feat_pd_names)

See the demos.py file for further examples. To run the demo code without previously adding the package:

  import sys
  sys.path.append(path_to_NEURAL_py)
  import demos


  demos.artefact_removal_examples()
  demos.generate_a_subset_of_features_example()
  demos.generate_all_features_example()
  demos.preprocessing_and_feature_extraction()
  demos.load_edf_preprocessing_and_feature_extraction()

All parameters are set the file utils.NEURAL_parameters.

Quantitative features

The feature set contains amplitude, spectral, connectivity, and burst annotation features. Amplitude features include range-EEG (D. O’ Reilly et al., 2012; see references), a clearly-defined alternative to amplitude-integrated EEG (aEEG). All features are generated for four different frequency bands (typically 0.5–4, 4–7, 7–13, and 13–30 Hz), with some exceptions. The following table describes the features in more detail:

feature name	description	FB
spectral_power	spectral power: absolute	yes
spectral_relative_power	spectral power: relative (normalised to total spectral power)	yes
spectral_flatness	spectral entropy: Wiener (measure of spectral flatness)	yes
spectral_entropy	spectral entropy: Shannon	yes
spectral_diff	difference between consecutive short-time spectral estimates	yes
spectral_edge_frequency	cut-off frequency (fc): 95% of spectral power contained between 0.5 and fc Hz	no
FD	fractal dimension	yes
amplitude_total_power	time-domain signal: total power	yes
amplitude_SD	time-domain signal: standard deviation	yes
amplitude_skew	time-domain signal: skewness (absolute value)	yes
amplitude_kurtosis	time-domain signal: kurtosis	yes
amplitude_env_mean	envelope: mean value	yes
amplitude_env_SD	envelope: standard deviation (SD)	yes
connectivity_BSI	brain symmetry index (see Van Putten 2007)	yes
connectivity_corr	correlation (Spearman) between envelopes of hemisphere-paired channels	yes
connectivity_coh_mean	coherence: mean value	yes
connectivity_coh_max	coherence: maximum value	yes
connectivity_coh_freqmax	coherence: frequency of maximum value	yes
rEEG_mean	range EEG: mean	yes
rEEG_median	range EEG: median	yes
rEEG_lower_margin	range EEG: lower margin (5th percentile)	yes
rEEG_upper_margin	range EEG: upper margin (95th percentile)	yes
rEEG_width	range EEG: upper margin - lower margin	yes
rEEG_SD	range EEG: standard deviation	yes
rEEG_CV	range EEG: coefficient of variation	yes
rEEG_asymmetry	range EEG: measure of skew about median	yes
IBI_length_max	burst annotation: maximum (95th percentile) inter-burst interval	no
IBI_length_median	burst annotation: median inter-burst interval	no
IBI_burst_prc	burst annotation: burst percentage	no
IBI_burst_number	burst annotation: number of bursts	no

FB: features generated for each frequency band (FB)

Files

Some python files (.py files) have a description and an example in the header. To read this header, type help(filename) in the console after importing (import filename). Directory structure is as follows:

├── amplitude_features.py                  # amplitude features
├── connectivity_features.py               # hemisphere connectivity features
├── demos.py                               # python file to give examples on how to use the code
├── fd_features.py                         # fractal dimension features
├── generate_all_features.py               # main function: generates feature set on EEG
├── IBI_features.py                        # features from the burst annotations  
├── LICENSE.md                             # license file  
├── mfiltfilt.py                           # MATLAB implementation of filtfilt function 
├── preprocessing_EEG.py                   # loads EEG from EDF files (including artefact removal)
├── README.md                              # readme file describing project
├── rEEG.py                                # range-EEG (similar to aEEG)
├── spectral_features.py                   # spectral features
└── utils.py                               # misc. functions

Test computer setup

• hardware: Intel Core i7-8700K @ 3.2GHz; 32GB memory.
• operating system: Windows 10 64-bit
• software: python 3.7

Licence

Copyright (c) 2020, Brian M. Murphy and John M. O' Toole, University College Cork
All rights reserved.

Redistribution and use in source and binary forms, with or without modification,
are permitted provided that the following conditions are met:

  Redistributions of source code must retain the above copyright notice, this
  list of conditions and the following disclaimer.

  Redistributions in binary form must reproduce the above copyright notice, this
  list of conditions and the following disclaimer in the documentation and/or
  other materials provided with the distribution.

  Neither the name of the University College Cork nor the names of its
  contributors may be used to endorse or promote products derived from
  this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR
ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

References

1. JM O’Toole and GB Boylan (2017). NEURAL: quantitative features for newborn EEG using Matlab. ArXiv e-prints, arXiv:1704.05694.

2. D O’Reilly, MA Navakatikyan, M Filip, D Greene, & LJ Van Marter (2012). Peak-to-peak amplitude in neonatal brain monitoring of premature infants. Clinical Neurophysiology, 123(11):2139–2153.

3. MJAM van Putten (2007). The revised brain symmetry index. Clinical Neurophysiology, 118(11):2362–2367.

4. T Higuchi (1988). Approach to an irregular time series on the basis of the fractal theory, Physica D: Nonlinear Phenomena, 31:277–283.

5. MJ Katz (1988). Fractals and the analysis of waveforms. Computers in Biology and Medicine, 18(3):145–156.

6. AV Oppenheim, RW Schafer. Discrete-Time Signal Processing. Prentice-Hall, Englewood Cliffs, NJ 07458, 1999.

7. JM O’ Toole, GB Boylan, S Vanhatalo, NJ Stevenson (2016). Estimating functional brain maturity in very and extremely preterm neonates using automated analysis of the electroencephalogram. Clinical Neurophysiology, 127(8):2910–2918

8. JM O’ Toole, GB Boylan, RO Lloyd, RM Goulding, S Vanhatalo, NJ Stevenson (2017). Detecting Bursts in the EEG of Very and Extremely Premature Infants using a Multi-Feature Approach. Medical Engineering and Physics, vol. 45, pp. 42-50, 2017. DOI:10.1016/j.medengphy.2017.04.003

9. JM O'Toole and GB Geraldine (2019). Quantitative Preterm EEG Analysis: The Need for Caution in Using Modern Data Science Techniques. Frontiers in Pediatrics 7, 174 DOI:10.3389/fped.2019.00174

Contact

Brian M. Murphy

Neonatal Brain Research Group,
INFANT Research Centre,
Department of Paediatrics and Child Health,
Room 2.18 UCC Academic Paediatric Unit, Cork University Hospital,
University College Cork,
Ireland

• email: Brian.M.Murphy AT umail dot ucc dot ie