# A Neonatal EEG Dataset Graded for Severity of Background Abnormalities The dataset consists of 169 multichannel EEG files of 1-hour in duration, recorded from 53 full-term newborns in the neonatal intensive care unit of the Cork University Maternity Hospital, Ireland. All 53 infants had received a diagnosis of hypoxic-ischaemic encephalopathy. The study to record the EEG was approved by the Cork Research Ethics Committee of the Cork Teaching Hospitals. Neonates were enrolled in the study after obtaining written and informed consent from a guardian or parent. The Cork Research Ethics Committee approved the publication of this fully-anonymised data set. Each 1-hour EEG was graded for severity of background abnormalities. Two experts in neonatal EEG graded each epoch independently. When grades differed between the experts, they jointly reviewed the EEG and agreed on a consensus grade. The grading system, detailed in [reference 1](##references), assesses EEG attributes such as amplitude and frequency, continuity, sleep--wake cycling, symmetry and synchrony, and abnormal waveforms. Four grades were used: normal or mildly abnormal (grade 1), moderately abnormal (grade 2), severely abnormal (grade 3), and inactive (grade 4). The EEG data could be used to develop automated grading algorithms (e.g. see [reference 2](##references)) or to assist in training for the review of background neonatal EEG. ## EEG file formats The EEG data is provided in 2 formats: European data format (EDF) files and as compressed comma separated variable (CSV) files. Each 1-hour epoch is stored as a separate file, using the file name convention `IDXX_epochY`. For example, file `ID10_epoch2` is the 2nd epoch for baby 10. All IDs are anonymised. The EDF files are stored in the `EDF_format/` folder and the CSV files are stored in the `CSV_format/` folder. The grade associated with each file is stored in the CSV `eeg_grades.csv` file. EDF format is a standard open format that most EEG viewers can read, for example the [EDFbrowser](https://www.teuniz.net/edfbrowser/) viewer can display EDF files. The CSV files are compressed using the [XZ](https://tukaani.org/xz/) compression format. ## Importing the data in programming environments The CSV files are the easiest format to import the EEG data. The following provides examples for file `filename.csv.xz` in [Python](https://www.python.org/), [R](https://www.r-project.org/), [Matlab](https://www.mathworks.com/products/matlab.html), and [Julia](https://julialang.org/) In Python and R, the compressed file can be read in directly. For example, in Python as ```python import pandas as pd eeg_df = pd.read_csv("filename.csv.xz") ``` and in `R` as ```R eeg_df <- read.csv('filename.csv.xz') ``` In Matlab and Julia, the file must be uncompressed first. Then the CSV file is imported. For example, in Matlab: ```matlab eeg_tb = readtable("filename.csv"); ``` where `filename.csv` is the uncompressed version of `filename.csv.xz`. In Julia, ```julia using CSV using DataFrames eeg_df = CSV.read("filename.csv", DataFrame) ``` ## References 1. DM Murray, GB Boylan, CA Ryan & S Connolly (2009) 'Early EEG findings in hypoxic-ischemic encephalopathy predict outcomes at 2 years.', Pediatrics 124, e459--67 [DOI:10.1542/peds.2008-2190](https://doi.org/10.1542/peds.2008-2190) 2. SA Raurale, GB Boylan, SR Mathieson, WP Marnane & JM O'Toole (2021) 'Grading hypoxic-ischemic encephalopathy in neonatal EEG with convolutional neural networks and quadratic time–frequency distributions', Journal of Neural Engineering 18, 046007 [DOI:10.1088/1741-2552/abe8ae](https://doi.org/10.1088/1741-2552/abe8ae) ## Contact John M O'Toole, Sean R Mathieson, Fabio Magarelli, William P Marnane, Gordon Lightbody, Geraldine B Boylan [INFANT Research Centre](https://www.infantcentre.ie/) at [University College Cork](https://www.ucc.ie/en/), Paediatric Academic Unit, Cork University Hospital, Wilton, Cork, Ireland Tel: +353 (0)21 4205082 Email: infant @ ucc .ie ___ This work is licensed under a [Creative Commons Attribution 4.0 International License](http://creativecommons.org/licenses/by/4.0/)