This database, collected at the Children’s Hospital Boston, consists of EEG recordings from pediatric subjects with intractable seizures. Subjects were monitored for up to several days following withdrawal of anti-seizure medication in order to characterize their seizures and assess their candidacy for surgical intervention. The recordings are grouped into 23 cases and were collected from 22 subjects (5 males, ages 3–22; and 17 females, ages 1.5–19).

Univ. of Bonn’ and ‘CHB-MIT Scalp EEG Database’ are publically available datasets which are the most sought after amongst researchers. Bonn dataset is very small compared to CHB-MIT. But still researchers prefer Bonn as it is in simple '.txt' format. The dataset being published here is a preprocessed form of CHB-MIT. The dataset is available in '.csv' format.

Artificial intelligence (AI) based automated epilepsy diagnosis has aimed to ease the burden of manual detection, prediction, and management of seizure and epilepsy-specific EEG signals for medical specialists. With increasing open-source, raw, and large EEG datasets, there is a need for data standardization of patient and seizure-sensitive AI analysis with reduced redundant information. This work releases a balanced, annotated, fixed time and length meta-data of CHB-MIT Scalp EEG database v1.0.0.0.

The work releases patient-specific (inter and intra) and patient non-specific EEG data extracted using specific time stamps of ictal, pre-ictal, post-ictal, peri-ictal, and non-seizure EEG provided in the original dataset (annotations). Further details of this metadata can be found in the provided csv file (CHB-MIT DB timestamp.csv). The released EEG data is available in csv format and class labels are provided in the last row of the csv files. Data of ch06, ch12, ch23, and ch24 in patient-specific and chb24_11 in patient non-specific have not been included. The importance of peri-ictal EEGs has been elucidated in Handa, P., & Goel, N. (2021). Peri‐ictal and non‐seizure EEG event detection using generated metadata. Expert Systems, e12929.

Latest update: The dataset has been moved to: Data of Meta-EEGs

Please only download and cite the latest version!

Compiled the original 2010 database from Physionet.org

This dataset only includes the waveforms from the original dataset that have seizure events. Seizure events are annotated in seizure_events.csv in seconds.

Note: the waveforms are taken in 256Hz, the event onset and offset times are denoted in seconds

The original dataset description:

This database, collected at the Children’s Hospital Boston, consists of EEG recordings from pediatric subjects with intractable seizures. Subjects were monitored for up to several days following withdrawal of anti-seizure medication in order to characterize their seizures and assess their candidacy for surgical intervention. Recordings, grouped into 23 cases, were collected from 22 subjects (5 males, ages 3–22; and 17 females, ages 1.5–19). (Case chb21 was obtained 1.5 years after case chb01, from the same female subject.) The file SUBJECT-INFO contains the gender and age of each subject. (Case chb24 was added to this collection in December 2010, and is not currently included in SUBJECT-INFO.)

Each case (chb01, chb02, etc.) contains between 9 and 42 continuous .edf files from a single subject. Hardware limitations resulted in gaps between consecutively-numbered .edf files, during which the signals were not recorded; in most cases, the gaps are 10 seconds or less, but occasionally there are much longer gaps. In order to protect the privacy of the subjects, all protected health information (PHI) in the original .edf files has been replaced with surrogate information in the files provided here. Dates in the original .edf files have been replaced by surrogate dates, but the time relationships between the individual files belonging to each case have been preserved. In most cases, the .edf files contain exactly one hour of digitized EEG signals, although those belonging to case chb10 are two hours long, and those belonging to cases chb04, chb06, chb07, chb09, and chb23 are four hours long; occasionally, files in which seizures are recorded are shorter.

All signals were sampled at 256 samples per second with 16-bit resolution. Most files contain 23 EEG signals (24 or 26 in a few cases). The International 10-20 system of EEG electrode positions and nomenclature was used for these recordings. In a few records, other signals are also recorded, such as an ECG signal in the last 36 files belonging to case chb04 and a vagal nerve stimulus (VNS) signal in the last 18 files belonging to case chb09. In some cases, up to 5 “dummy” signals (named "-") were interspersed among the EEG signals to obtain an easy-to-read display format; these dummy signals can be ignored.

The file RECORDS contains a list of all 664 .edf files included in this collection, and the file RECORDS-WITH-SEIZURES lists the 129 of those files that contain one or more seizures. In all, these records include 198 seizures (182 in the original set of 23 cases); the beginning ([) and end (]) of each seizure is annotated in the .seizure annotation files that accompany each of the files listed in RECORDS-WITH-SEIZURES. In addition, the files named chbnn-summary.txt contain information about the montage used for each recording, and the elapsed time in seconds from the beginning of each .edf file to the beginning and end of each seizure contained in it.

This database is described in:

Ali Shoeb. Application of Machine Learning to Epileptic Seizure Onset Detection and Treatment. PhD Thesis, Massachusetts Institute of Technology, September 2009.

Please cite this publication when referencing this material, and also include the standard citation for PhysioNet:

Goldberger AL, Amaral LAN, Glass L, Hausdorff JM, Ivanov PCh, Mark RG, Mietus JE, Moody GB, Peng C-K, Stanley HE. PhysioBank, PhysioToolkit, and PhysioNet: Components of a New Research Resource for Complex Physiologic Signals. Circulation 101(23):e215-e220 [Circulation Electronic Pages; http://circ.ahajournals.org/cgi/content/full/101/23/e215]; 2000 (June 13)."

Recordings, grouped into 23 cases, were collected from 22 subjects (5 males, ages 3–22; and 17 females, ages 1.5–19). (Case chb21 was obtained 1.5 years after case chb01, from the same female subject.) The file SUBJECT-INFO contains the gender and age of each subject. (Case chb24 was added to this collection in December 2010, and is not currently included in SUBJECT-INFO.)

Each case (chb01, chb02, etc.) contains between 9 and 42 continuous .edf files from a single subject. Hardware limitations resulted in gaps between consecutively-numbered .edf files, during which the signals were not recorded; in most cases, the gaps are 10 seconds or less, but occasionally there are much longer gaps. In order to protect the privacy of the subjects, all protected health information (PHI) in the original .edf files has been replaced with surrogate information in the files provided here. Dates in the original .edf files have been replaced by surrogate dates, but the time relationships between the individual files belonging to each case have been preserved. In most cases, the .edf files contain exactly one hour of digitized EEG signals, although those belonging to case chb10 are two hours long, and those belonging to cases chb04, chb06, chb07, chb09, and chb23 are four hours long; occasionally, files in which seizures are recorded are shorter.

All signals were sampled at 256 samples per second with 16-bit resolution. Most files contain 23 EEG signals (24 or 26 in a few cases). The International 10-20 system of EEG electrode positions and nomenclature was used for these recordings. In a few records, other signals are also recorded, such as an ECG signal in the last 36 files belonging to case chb04 and a vagal nerve stimulus (VNS) signal in the last 18 files belonging to case chb09. In some cases, up to 5 “dummy” signals (named "-") were interspersed among the EEG signals to obtain an easy-to-read display format; these dummy signals can be ignored.

The file RECORDS contains a list of all 664 .edf files included in this collection, and the file RECORDS-WITH-SEIZURES lists the 129 of those files that contain one or more seizures. In all, these records include 198 seizures (182 in the original set of 23 cases); the beginning ([) and end (]) of each seizure is annotated in the .seizure annotation files that accompany each of the files listed in RECORDS-WITH-SEIZURES. In addition, the files named chbnn-summary.txt contain information about the montage used for each recording, and the elapsed time in seconds from the beginning of each .edf file to the beginning and end of each seizure contained in it.

THIS RESOURCE IS NO LONGER IN SERVICE. Documented on November 22, 2022. Data set collected at the Children''s Hospital Boston, of EEG recordings from pediatric subjects with intractable seizures. Subjects were monitored for up to several days following withdrawal of anti-seizure medication in order to characterize their seizures and assess their candidacy for surgical intervention. All signals were sampled at 256 samples per second with 16-bit resolution. Most files contain 23 EEG signals (24 or 26 in a few cases).

Annotated EDF file for CHB-MIT Scalp EEG Database

Recordings, grouped into 23 cases, were collected from 22 subjects (5 males, ages 3–22; and 17 females, ages 1.5–19). (Case chb21 was obtained 1.5 years after case chb01, from the same female subject.) The file SUBJECT-INFO contains the gender and age of each subject. (Case chb24 was added to this collection in December 2010, and is not currently included in SUBJECT-INFO.)

Each case (chb01, chb02, etc.) contains between 9 and 42 continuous .edf files from a single subject. Hardware limitations resulted in gaps between consecutively-numbered .edf files, during which the signals were not recorded; in most cases, the gaps are 10 seconds or less, but occasionally there are much longer gaps. In order to protect the privacy of the subjects, all protected health information (PHI) in the original .edf files has been replaced with surrogate information in the files provided here. Dates in the original .edf files have been replaced by surrogate dates, but the time relationships between the individual files belonging to each case have been preserved. In most cases, the .edf files contain exactly one hour of digitized EEG signals, although those belonging to case chb10 are two hours long, and those belonging to cases chb04, chb06, chb07, chb09, and chb23 are four hours long; occasionally, files in which seizures are recorded are shorter.

All signals were sampled at 256 samples per second with 16-bit resolution. Most files contain 23 EEG signals (24 or 26 in a few cases). The International 10-20 system of EEG electrode positions and nomenclature was used for these recordings. In a few records, other signals are also recorded, such as an ECG signal in the last 36 files belonging to case chb04 and a vagal nerve stimulus (VNS) signal in the last 18 files belonging to case chb09. In some cases, up to 5 “dummy” signals (named "-") were interspersed among the EEG signals to obtain an easy-to-read display format; these dummy signals can be ignored.

The file RECORDS contains a list of all 664 .edf files included in this collection, and the file RECORDS-WITH-SEIZURES lists the 129 of those files that contain one or more seizures. In all, these records include 198 seizures (182 in the original set of 23 cases); the beginning ([) and end (]) of each seizure is annotated in the .seizure annotation files that accompany each of the files listed in RECORDS-WITH-SEIZURES. In addition, the files named chbnn-summary.txt contain information about the montage used for each recording, and the elapsed time in seconds from the beginning of each .edf file to the beginning and end of each seizure contained in it.

CHB MIT Scalp EEG Database Expo

## Overview

CHB MIT Scalp EEG Database Expo is a dataset for classification tasks - it contains ObjectsCHB MIT Scalp EEG Databas annotations for 2,531 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [Public Domain license](https://creativecommons.org/licenses/Public Domain).

Artificial intelligence (AI) based automated epilepsy diagnosis has aimed to ease the burden of manual detection, prediction, and management of seizure and epilepsy-specific EEG signals for medical specialists. Existing research work in this domain has highlighted the significance of 2D EEG frames extracted through various processing pipelines over 1D signal analysis using various CNN architectures like AlexNet, LeNet. This is a pre-processed image (rhythmicity spectrogram) dataset generated from the CHB-MIT EEG scalp database. The dataset consists of 105 frames from chb01, 30 frames from chb02, 90 frames from chb05, and 75 frames from chb05 separately from both ictal and non-seizure edf files. The total image frames and ictal time (20 ictal signals) are 600 frames and 25 minutes respectively. The dataset has been divided into train, test and validate folders wherein seizure and non-seizure EEG images have been put in png format. It can be incorporated in the machine and deep learning pipelines for the detection of seizure and non-seizure EEG images.

For further technical details see the following publication: Handa, P., & Goel, N. (2021, August). Epileptic Seizure Detection Using Rhythmicity Spectrogram and Cross-Patient Test Set. In 2021 8th International Conference on Signal Processing and Integrated Networks (SPIN) (pp. 898-902). IEEE.

Dataset

This dataset was created by ImenJmal

Contents

Changes in the frequency composition of the human electroencephalogram are associated with the transitions to epileptic seizures. Cross-frequency coupling (CFC) is a measure of neural oscillations in different frequency bands and brain areas, and specifically phase–amplitude coupling (PAC), a form of CFC, can be used to characterize these dynamic transitions. In this study, we propose a method for seizure detection and prediction based on frequency domain analysis and PAC combined with machine learning. We analyzed two databases, the Siena Scalp EEG database and the CHB-MIT database, and used the frequency features and modulation index (MI) for time-dependent quantification. The extracted features were fed to a random forest classifier for classification and prediction. The seizure prediction horizon (SPH) was also analyzed based on the highest-performing band to maximize the time for intervention and treatment while ensuring the accuracy of the prediction. Under comprehensive consideration, the results demonstrate that better performance could be achieved at an interval length of 5 min with an average accuracy of 85.71% and 95.87% for the Siena Scalp EEG database and the CHB-MIT database, respectively. As for the adult database, the combination of PAC analysis and classification can be of significant help for seizure detection and prediction. It suggests that the rarely used SPH also has a major impact on seizure detection and prediction and further explorations for the application of PAC are needed.

IntroductionLong-term electroencephalography (EEG) monitoring is advised to patients with refractory epilepsy who have a failure of anti-seizure medication and therapy. However, its real-life application is limited mainly due to the use of multiple EEG channels. We proposed a patient-specific deep learning-based single-channel seizure detection approach using the long-term scalp EEG recordings of the Children’s Hospital Boston-Massachusetts Institute of Technology (CHB-MIT) dataset, in conjunction with neurologists’ confirmation of spatial seizure characteristics of individual patients.MethodsWe constructed 18-, 4-, and single-channel seizure detectors for 13 patients. Neurologists selected a specific channel among four channels, two close to the behind-the-ear and two at the forehead for each patient, after reviewing the patient’s distinctive seizure locations with seizure re-annotation.ResultsOur multi- and single-channel detectors achieved an average sensitivity of 97.05–100%, false alarm rate of 0.22–0.40/h, and latency of 2.1–3.4 s for identification of seizures in continuous EEG recordings. The results demonstrated that seizure detection performance of our single-channel approach was comparable to that of our multi-channel ones.DiscussionWe suggest that our single-channel approach in conjunction with clinical designation of the most prominent seizure locations has a high potential for wearable seizure detection on long-term EEG recordings for patients with refractory epilepsy.

A dataset created from EEG signals from the CHB-MIT Scalp EEG Database, specifically from Patient 15.

Using the raw EEG signals, windows of 30 seconds (7680 samples each) were randomly sampled, converted to spectrograms with 1 sec (256 samples) in each segment using scipy's signal.stft function.

Only train and validation sets are included. The training samples are sourced from recordings that were recorded earlier than validation samples.

The {set}_X.npy are files containing the numpy arrays for multichannel spectrograms and have a shape of n x f x t x c where: * n = number of samples in the set * f = frequencies (spectral axis) * t = time (temporal axis) * c = number of channels

The {set}_y.npy are files containing the target for each spectrogram. It is a continuous measure representing the number of seconds from the most recent sample in the window to the next epilepsy event.

This dataset was created to be used in a regression task. However, y can easily be converted to binary or a multivariate variable and be used for classification using specified thresholds.

A dataset created from EEG signals from the CHB-MIT Scalp EEG Database, specifically from Patient 1.

Using the raw EEG signals, windows of 30 seconds (7680 samples each) were randomly sampled.

Train, validation, and test sets are included. The training samples are sourced from recordings that were recorded earlier than validation samples which were recorder earlier than test samples.

The {set}_X.npy are files containing the numpy arrays for multichannel signals and have a shape of n x w x c where: * n = number of samples in the set * w = number of signal sample in a window sample * c = number of channels

The {set}_y.npy are files containing the target for each window. It is a continuous measure representing the number of seconds from the most recent sample in the window to the next epilepsy event.

This dataset was created to be used in a regression task. However, y can easily be converted to binary or a multivariate variable and be used for classification using specified thresholds.

CNN has demonstrated remarkable performance in EEG signal detection, yet it still faces limitations in terms of global perception. Additionally, due to individual differences in EEG signals, the generalization ability of epilepsy detection models is week. To address this issue, this paper presents a cross-patient epilepsy detection method utilizing a multi-head self-attention mechanism. This method first utilizes Short-Time Fourier Transform (STFT) to transform the original EEG signals into time-frequency features, then models local information using Convolutional Neural Network (CNN), subsequently captures global dependency relationships between features using the multi-head self-attention mechanism of Transformer, and finally performs epilepsy detection using these features. Meanwhile, this model employs a light multi-head attention mechanism module with an alternating structure, which can comprehensively extract multi-scale features while significantly reducing computational costs. Experimental results on the CHB-MIT dataset show that the proposed model achieves accuracy, sensitivity, specificity, F1 score, and AUC of 92.89%, 96.17%, 92.99%, 94.41%, and 96.77%, respectively. Compared to the existing methods, the method proposed in this paper obtains better performance along with better generalization.

Source: PhysioNet Original Authors: Ali Shoeb, John Guttag, and collaborators from Children’s Hospital Boston (CHB) and MIT Original Publication Date: June 9, 2010 Original DOI: https://doi.org/10.13026/C2K01R

Dataset Overview

The CHB-MIT Scalp EEG Database is a collection of EEG recordings from pediatric subjects with intractable seizures. The recordings were collected at Children’s Hospital Boston to help understand seizure patterns and evaluate patients for potential surgical treatment. This dataset includes annotated seizure onsets and ends across 23 cases from 22 patients (ages 1.5 to 22 years).

Each case consists of multiple continuous EEG recordings saved in .edf format, with most files covering a duration of one hour and sampled at 256 Hz. Cases include EEG channels following the International 10-20 electrode placement system, along with additional signals such as ECG and vagal nerve stimulus in some cases.

Data Format

EDF Files: 664 .edf files organized by case (e.g., chb01, chb02), representing continuous EEG recordings. Annotations: Seizure onset and end times are annotated within the files. Signal Sampling: All signals are sampled at 256 Hz with 16-bit resolution, including 23 EEG channels in most cases. Potential Uses This dataset is valuable for research in EEG signal analysis and seizure detection, including patient-specific seizure onset detection through machine learning.

Data Structure

Recordings: Organized into 23 cases across 664 .edf files, with 129 files containing seizure annotations. Annotations: Each seizure's start and end points are annotated, and the accompanying files detail the time elapsed to seizure onset and offset within each recording. Subject Info: Includes demographic information like age and gender for each case. Additional Signals: Some cases contain ECG or vagal nerve stimulus (VNS) signals. Potential Uses This dataset is suitable for developing and evaluating machine learning models for seizure detection, as demonstrated in research by Ali Shoeb, John Guttag, and colleagues. The annotated seizure events provide a labeled dataset valuable for testing seizure onset prediction algorithms and could contribute to advancements in real-time seizure monitoring devices.

Citation

If you use this dataset, please cite it in accordance with the original authors’ guidelines:

Database Citation:

Guttag, J. (2010). CHB-MIT Scalp EEG Database (version 1.0.0). PhysioNet. Available at https://doi.org/10.13026/C2K01R

Research Work

Ali Shoeb. Application of Machine Learning to Epileptic Seizure Onset Detection and Treatment. PhD Thesis, Massachusetts Institute of Technology, September 2009.

PhysioNet Standard Citation

Goldberger, A., Amaral, L., Glass, L., et al. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation, 101 (23), e215–e220.

Acknowledgments

This database was developed by researchers at Children’s Hospital Boston (CHB) and the Massachusetts Institute of Technology (MIT), including contributions from clinical investigators Jack Connolly, Herman Edwards, Blaise Bourgeois, and S. Ted Treves, as well as researchers Ali Shoeb and Professor John Guttag.

Disclaimer

This dataset was originally made available on PhysioNet for research purposes. For additional information and official access, please refer to the PhysioNet website at https://doi.org/10.13026/C2K01R. This version is intended solely for research use on Kaggle and is not an official re-release by the original authors.

A dataset created from EEG signals from the CHB-MIT Scalp EEG Database, specifically from Patient 1.

Using the raw EEG signals, windows of 30 seconds (7680 samples each) were randomly sampled, converted to spectrograms with 1 sec (256 samples) in each segment using scipy's signal.stft function.

Train, validation, and test sets are included. The training samples are sourced from recordings that were recorded earlier than validation samples which were recorder earlier than test samples.

The {set}_X.npy are files containing the numpy arrays for multichannel spectrograms and have a shape of n x f x t x c where: * n = number of samples in the set * f = frequencies (spectral axis) * t = time (temporal axis) * c = number of channels

The {set}_y.npy are files containing the target for each spectrogram. It is a continuous measure representing the number of seconds from the most recent sample in the window to the next epilepsy event.

This dataset was created to be used in a regression task. However, y can easily be converted to binary or a multivariate variable and be used for classification using specified thresholds.