CHB-MIT

Introduction

The CHB-MIT Scalp EEG Database consists of EEG recordings from pediatric subjects with intractable seizures. This dataset was collected at the Children's Hospital Boston and includes recordings from 22 subjects (5 males, ages 3-22; and 17 females, ages 1.5-19) with epilepsy. The recordings contain 198 annotated seizures and were originally collected to characterize seizures and assess patients' candidacy for surgical intervention.

Overview of the experiment

Subjects were monitored for up to several days following withdrawal of anti-seizure medication in a controlled hospital environment. The purpose was to capture and characterize their seizure patterns using continuous scalp EEG monitoring. Each case (subject) contains between 9 and 42 continuous EEG recording files. All signals were sampled at 256 samples per second with 16-bit resolution. Most files contain 23 EEG signals recorded using the International 10-20 system of EEG electrode positions and nomenclature. The recordings use bipolar montages, where each channel represents the potential difference between two electrode sites. Hardware limitations resulted in gaps between consecutively-numbered files, typically 10 seconds or less, during which signals were not recorded. Most recording files contain exactly one hour of digitized EEG signals, though some cases contain two-hour or four-hour recordings. Additional signals such as ECG and vagal nerve stimulus (VNS) were recorded in some cases.

Description of the preprocessing if any

The original .edf files from PhysioNet have been converted to BIDS format. Channel names have been standardized to match the standard 10-05 montage naming convention. Bipolar channel pairs are represented in the format "Electrode1-Electrode2" (e.g., "FP1-F7"). Non-EEG channels such as ECG are preserved with appropriate BIDS channel types. Channels that did not match expected formats or could not be mapped to the standard montage were marked as "misc" type. All protected health information (PHI) in the original files has been replaced with surrogate information. Dates have been replaced with surrogate dates while preserving time relationships between files. Subject birthdates are calculated based on age at recording time when available.

Description of the event values if any

The events.tsv files contain seizure onset and offset annotations. Each seizure event has:

• onset: Time in seconds from the beginning of the recording when the seizure starts
• duration: Duration of the seizure in seconds
• value: "seizure" - indicating a seizure event
• sample: Sample number at onset

The seizure annotations were originally marked with '[' for onset and ']' for offset in the .seizures annotation files and have been converted to BIDS-compliant event format. In total, the dataset contains 198 seizure events across all subjects (182 in the original 23 cases, plus 16 additional seizures from case chb24 added in December 2010).

Citation

When using this dataset, please cite:

1. Ali Shoeb. Application of Machine Learning to Epileptic Seizure Onset Detection and Treatment. PhD Thesis, Massachusetts Institute of Technology, September 2009. http://hdl.handle.net/1721.1/54669

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

1. Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., ... & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220.

Data curators: Pierre Guetschel (BIDS conversion)

Original data collection team:

• Jack Connolly, REEGT (Children's Hospital Boston)
• Herman Edwards, REEGT (Children's Hospital Boston)
• Blaise Bourgeois, MD (Children's Hospital Boston)
• S. Ted Treves, MD (Children's Hospital Boston)
• Ali Shoeb, PhD (Massachusetts Institute of Technology)
• Professor John Guttag (Massachusetts Institute of Technology)

---

Automatic report

Report automatically generated by mnebids.makereport().

> The CHB-MIT dataset was created by Jack Connolly, Herman Edwards, Blaise Bourgeois, S. Ted Treves, Ali Shoeb, and John Guttag and conforms to BIDS version 1.7.0. This report was generated with MNE-BIDS (https://doi.org/10.21105/joss.01896). The dataset consists of 24 participants (comprised of 5 male and 18 female participants; handedness were all unknown; ages ranged from 71.0 to 91.0 (mean = 79.04, std = 5.51; 1 with unknown age)) . Data was recorded using an EEG system sampled at 256.0 Hz with line noise at n/a Hz. There were 686 scans in total. Recording durations ranged from 600.0 to 14427.0 seconds (mean = 5158.26, std = 3657.58), for a total of 3538564.32 seconds of data recorded over all scans. For each dataset, there were on average 26.03 (std = 3.81) recording channels per scan, out of which 26.03 (std = 3.81) were used in analysis (0.0 +/- 0.0 were removed from analysis).