Leipzig Heart Center ECG-Database: Arrhythmias in Children and Patients with Congenital Heart Disease

Abstract

Interpretation of Electrocardiograms (ECG) is increasingly complemented by algorithms. These algorithms are based on large datasets. This ECG database consists of children and adults with congenital heart defects (CHD) including many arrhythmia annotations. This dataset, notable for its manual annotations and inclusion of intracardiac electrograms alongside traditional 12-lead ECGs, offers 1075.85 minutes of recordings (in total 113924 annotated beats) that capture various cardiac rhythms and arrhythmias such as supraventricular tachycardia and ventricular tachycardia. The data were meticulously collected from patients undergoing electrophysiological studies, with subsequent annotations by expert reviewers using the LightWAVE® software. The significance of this database lies in its focus on pediatric arrhythmias and arrhythmias of patients with congenital heart defect areas currently underrepresented in existing datasets, which predominantly feature adult pathologies. This resource aims to enhance algorithmic development for ECG interpretation, leveraging machine learning to improve diagnosis and treatment outcomes in these sensitive groups. The dataset not only serves as a critical tool for developing precision medicine but also sets a precedent for future expansions to include a broader spectrum of congenital heart defects conditions, thereby supporting the evolution of cardiac care through advanced computational techniques.

Background

Electrocardiography (ECG) is the most important diagnostic tool, in diagnosing cardiac arrhythmias. Interpretation of ECG is increasingly complemented by algorithms based on large datasets. So far these ECG-datasets mainly contain ECGs of adults, and the focus is primarily on adult pathologies like myocardial infarction, atrial fibrillation and conduction diseases [1-4]. The main arrhythmias of children, however, are completely different from arrhythmias of adults. The most common arrhythmias in children are supraventricular tachycardias like atrioventricular reentrant tachycardias and atrioventricular nodal reentrant tachycardias. Databases containing these arrhythmias are very rare. The aim of this study was to create an ECG-database of children with arrhythmia annotations beat per beat. Additionally to the arrhythmias of children, this is the first database containing also ECG data of patients with congenital heart disease. These patients have a high percentage of conduction anomalies (like right bundle branch block) and a risk for ventricular tachycardias. This resource aims to enhance algorithmic development for ECG interpretation, leveraging machine learning to improve diagnosis and treatment outcomes in children and patients with congenital heart disease.

Methods

The raw signal data were extracted from electrophysiological studies of children and patients with congenital heart defects.

Signals were recorded using CardioLab^® (GE Medical Systems, Milwaukee, WI). Data were exported as a .txt format. The dataset is provided in the WaveForm Database (WFDB) format. The WFDB Software package was used to convert the data into the WFDB format [5]. The WFDB Software package was also used for various processing steps and validation in the annotation process of the data. The recordings were digitized at 977 samples per second. Filter settings were 0.05 Hz-100 Hz. The signals from the individual leads are the raw data as exported by CardioLab^®. No filters or other signal quality enhancements have been applied afterward.

To annotate the dataset, the Software LightWAVE^® was used [6, 7]. The annotations were manually annotated by two ECG experts. The first expert annotated all ECGs. To support the process, the first expert could annotate long ranges of the same annotation by marking the start and end. By using the WFDB Software package and an R-peak detection algorithm, the ranges were filled [8]. The second expert checked all annotations and made corrections if necessary. For unclear parts of the record, further experts were consulted.

Data Description

The ECG dataset comprises detailed recordings from both children with supraventricular tachycardias and adultswith congenital heart defects, specifically categorized into separate groups. The files are systematically named to reflect the patient group: filenames x001 through x029 represent children's ECGs, while x100 through x109 denote adult patients' ECGs.

The files children-subject-info.csv and adults-subject-info.csv provide the following information:

• patient ID

• patient demographics (gender, age)

• patient diagnosis

• the location of the accessory pathway

• the duration of the ECG that is provided for the patient

The dataset consists of 39 ECGs of 39 patients (length ranging from 00:01:17 hours to 02:30:31 hours). In total 113924 beats were annotated. Normal sinus rhythm beats are the majority (33%), but the database contains also sinus rhythm beats with Preexcitation (12%) or complete right bundle branch block (11%). Supraventricular tachycardias (23%) are subdivided into atrioventricular reentrant tachycardias (15%) and atrioventricular nodal reentrant tachycardias ( 8%). 12% of all beats are paced beats (7% atrial paced beats and 5% ventricular paced beats) and 7% are premature beats, as premature ventricular beats, premature atrial beats and preexcited premature atrial beats.

The distribution of beat types annotated in this database is as follows:

*custom Aux Strings for standard WFDB annotations to further specify these annotations.

^#custom annotations used in addition to the standard WFDB annotations

This dataset includes the following types of rhythm annotations:

For information about the entire dataset and the abbreviations and annotations used in it, the file dataset_info.csv can be used.

Usage Notes

ECG datasets are crucial for developing AI-based arrhythmia detection algorithms. However, most existing datasets focus on adult pathologies, such as myocardial infarction, conduction diseases, or atrial fibrillation [1-4, 9]. Datasets including children are rare [2], and those featuring pediatric-specific conditions like supraventricular tachycardias are especially limited [1]. This manually annotated open-source database addresses this gap by providing normal and pathological ECGs of children, supporting advancements in AI-based rhythm analysis [10, 11].

This dataset provides ECG recordings with arrhythmia annotations, suitable for developing machine learning models to detect various cardiac conditions. The ECGs include:

• Supraventricular tachycardias (e.g., AVRT, AVNRT)

• Ventricular tachycardias

• Preexcitation syndrome

• Paced rhythms

• Complete right bundle branch block (CRBBB)

To utilize the data, researchers may use the WFDB Software Package for reading ECG data.

Each patient's data is encapsulated in three associated files:

1. Header File (.hea): This file provides metadata about the ECG recordings, including the number of channels, sampling frequency, and signal gain, essential for understanding and processing the ECG data correctly.

2. Data File (.dat): This file contains the actual ECG signal data, allowing detailed analysis and application of processing algorithms.

3. Annotation File (.atr): This file includes annotations of the ECG signals, detailing various cardiac events and rhythms identified by expert analysis, which are crucial for training and validating diagnostic algorithms.

Further details about how to use and how to get started with the dataset can be found in the README.md file.

As these ECGs are extracted from electrophysiological studies, they include segments with non-physiological stimulation. Currently, the database of adults with congenital heart defects is limited to patients with Tetralogy of Fallot, but plans are underway to include other congenital heart defects.

Ethics

Patients or their parents/legal guardians provided written informed consent prior to electrophysiological study. The study was approved by the institutional ethics committee (ethics approval number 120/23-ek) and fully complies with the Declaration of Helsinki.

Conflicts of Interest

The authors have no conflicts of interest to declare.

References

1. Zheng J, Zhang J, Danioko S, Yao H, Guo H, Rakovski C. A 12-lead electrocardiogram database for arrhythmia research covering more than 10,000 patients. Sci Data. 2020;7(1):48. https://doi.org/10.1038/s41597-020-0386-x
2. Wagner P, Strodthoff N, Bousseljot RD, Kreiseler D, Lunze FI, Samek W, et al. PTB-XL: A large publicly available ECG dataset. Sci Data. 2020;7(1):154. https://doi.org/10.1038/s41597-020-0495-6
3. Attia ZI, Noseworthy PA, Lopez-Jimenez F, Asirvatham SJ, Deshmukh AJ, Gersh BJ, et al. An artificial intelligence-enabled ECG algorithm for the identification of patients with atrial fibrillation during sinus rhythm: a retrospective analysis of outcome prediction. Lancet. 2019;394(10201):861–867. https://doi.org/10.1016/S0140-6736(19)31721-0
4. Yoo H, Yum Y, Park SW, Lee JM, Jang M, Kim Y, et al. Standardized database of 12-lead electrocardiograms with a common standard for the promotion of cardiovascular research: KURIAS-ECG. Healthc Inform Res. 2023;29(2):132–144. https://doi.org/10.4258/hir.2023.29.2.132
5. Xie C, McCullum L, Johnson A, Pollard T, Gow B, Moody B. Waveform Database Software Package (WFDB) for Python (version 4.1.0). PhysioNet. 2023. https://doi.org/10.13026/9njx-6322
6. Goldberger AL, Amaral LA, Glass L, Hausdorff JM, Ivanov PC, Mark RG, et al. PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals. Circulation. 2000;101(23):E215–220. https://doi.org/10.1161/01.cir.101.23.e215
7. Moody GB, Mark RG. The impact of the MIT-BIH arrhythmia database. IEEE Eng Med Biol Mag. 2001;20(3):45–50. https://doi.org/10.1109/51.932724
8. Kathirvel P, Sabarimalai Manikandan M, Prasanna SRM. An efficient R-peak detection based on new nonlinear transformation and first-order Gaussian differentiator. Cardiovasc Eng Technol. 2011;2(4):408–425. https://doi.org/10.1007/s13239-011-0065-3
9. Boulif A, Ananou B, Ouladsine M, Delliaux S. A literature review: ECG-based models for arrhythmia diagnosis using artificial intelligence techniques. Bioinform Biol Insights. 2023;17:11779322221149600. https://doi.org/10.1177/11779322221149600
10. Teich L, Franke D, Michaelis A, Dähnert I, Gebauer RA, Markel F, et al. Development of an AI-based automated analysis of pediatric Apple Watch iECGs. Front Pediatr. 2023;11:1185629. https://doi.org/10.3389/fped.2023.1185629
11. Yildirim O, Talo M, Ciaccio EJ, Tan RS, Acharya UR. Accurate deep neural network model to detect cardiac arrhythmia on more than 10,000 individual subject ECG records. Comput Methods Programs Biomed. 2020;197:105740. https://doi.org/10.1016/j.cmpb.2020.105740