Brugada-HUCA: 12-Lead ECG Recordings for the Study of Brugada Syndrome

@article{PhysioNet-brugada-huca-1.0.0,
  author = {{Costa Cortez}, Nahuel and {Garcia Iglesias}, Daniel},
  title = {{Brugada-HUCA: 12-Lead ECG Recordings for the Study of Brugada Syndrome}},
  journal = {{PhysioNet}},
  year = {2026},
  month = feb,
  note = {Version 1.0.0},
  doi = {10.13026/0m2w-dy83},
  url = {https://doi.org/10.13026/0m2w-dy83}
}

Abstract

We present a novel dataset of 12-lead electrocardiogram (ECG) recordings developed to support the study and classification of Brugada syndrome, a rare but potentially fatal cardiac arrhythmia. The data were collected retrospectively from patients evaluated at the Cardiology Department of the Hospital Universitario Central de Asturias (HUCA) and reviewed by clinical experts. Diagnostic labels were assigned according to internationally accepted criteria.

The dataset comprises 363 subjects in total, including 76 patients diagnosed with Brugada syndrome and 287 healthy control subjects. Each recording is accompanied by diagnostic information.

This resource aims to facilitate the development and validation of machine learning algorithms and computational tools for ECG interpretation, supporting improved early detection and clinical decision-making. The dataset, fully anonymized in accordance with ethical and privacy standards, provides a reproducible foundation for collaborative research into Brugada syndrome and related cardiac conditions.

Background

Brugada syndrome is a rare but potentially life-threatening cardiac arrhythmia characterized by distinctive ECG abnormalities, including a coved-type ST-segment elevation in leads V1–V3 often accompanied by a right bundle branch block pattern. Diagnosis is primarily based on ECG findings, either spontaneous or drug-induced, combined with clinical criteria such as syncope, documented ventricular arrhythmias, or a family history of sudden cardiac death [1, 2].

Despite clear diagnostic criteria, distinguishing Brugada syndrome from other ECG abnormalities remains challenging. Accurate identification is essential for risk stratification and prevention of sudden cardiac death [2]. Prior studies have explored quantitative and spectral ECG features to improve diagnostic accuracy and prognostic assessment [3, 4].

While several databases exist for general ECG analysis, few focus specifically on Brugada syndrome, and even fewer include expert-reviewed and labeled ECG recordings. This dataset therefore provides a unique, openly accessible resource for studying Brugada-specific ECG patterns and developing data-driven diagnostic algorithms, contributing to improved clinical interpretation and patient management.

Methods

A total of 363 standard 12-lead ECG recordings were collected from individuals evaluated for suspected Brugada syndrome at the Cardiology Department of the Hospital Universitario Central de Asturias (HUCA). Each recording has a duration of 12 seconds and a sampling frequency of 100 Hz.

Inclusion and Exclusion Criteria

Inclusion criteria comprised adult patients who underwent ECG evaluation as part of a clinical assessment for suspected Brugada syndrome or as part of routine cardiac screening. Exclusion criteria included poor-quality ECGs with significant noise or artifacts, incomplete lead information, or missing demographic or diagnostic data. Only recordings with adequate signal quality across all 12 leads were retained.

Dataset Composition

The final dataset consists of 363 ECG recordings, including 76 confirmed Brugada syndrome cases and 287 normal control recordings. Diagnostic classification (Normal vs. Brugada) followed internationally accepted consensus criteria for Brugada syndrome.

Annotation and Labeling Procedure

Diagnostic labeling was performed by at least one expert electrophysiologist with clinical experience in the diagnosis and management of Brugada syndrome. Each ECG was blindly reviewed, and labels were assigned based on the presence or absence of Brugada-type ECG patterns according to the established diagnostic standards. Ambiguous cases were discussed and adjudicated through consensus review among the clinical experts to ensure labeling consistency.

Preprocessing

Prior to dataset release, all recordings were reviewed to verify signal quality. No additional filtering or preprocessing (such as baseline wander removal, normalization, or resampling) was applied, allowing users to apply their own preferred preprocessing pipelines. All patient identifiers and protected health information were removed in accordance with institutional anonymization protocols and ethical guidelines.

Data Description

The dataset contains a total of 363 ECG waveform files and 363 corresponding metadata entries, ensuring a one-to-one correspondence between each ECG recording and its metadata record.

Each waveform file consists of 2 standard ECG leads — I, II, III, aVR, aVL, aVF, V1, V2, V3, V4, V5, and V6, recorded over a 12-second period at a sampling frequency of 100 Hz. Signal amplitudes are expressed in millivolts (mV) and conform to standard diagnostic ECG conventions.

Metadata File

A single comma-separated values (CSV) file accompanies the ECG recordings, containing per-subject metadata and diagnostic information. Each row corresponds to one recording, identified by a unique patient_id. The table below summarizes the variables included in the metadata file:

File Organization

Each ECG recording is provided in WFDB-compatible format, consisting of:

• A signal file (.dat) containing the raw waveform data.
• A header file (.hea) describing sampling frequency, signal format, and lead names.

The metadata CSV file is located in the root folder of the dataset. The filename convention ensures direct correspondence between waveform and metadata entries:

<patient_id>.dat / <patient_id>.hea → entry with patient_id in the metadata file.

Together, these files form a complete and standardized dataset ready for analysis using the WFDB Python toolbox, MATLAB, or other compatible signal processing environments.

Usage Notes

The dataset is distributed in WFDB-compatible format, consisting of paired .dat (signal) and .hea (header) files for each subject, along with a metadata.csv file describing diagnostic and clinical variables. The data can be easily accessed and analyzed using the open-source WFDB Python library, which provides functions for reading, visualizing, and processing physiological signal data.

Loading the Dataset

Reading Metadata

The metadata file (metadata.csv) provides diagnostic labels and auxiliary information for each subject. The following example demonstrates how to load and explore the metadata using Python:

import pandas as pd

# Load metadata
metadata = pd.read_csv('metadata.csv')

# Display basic statistics
print(metadata.head())
print(f"Total subjects: {len(metadata)}")
print(f"Brugada patients: {(metadata['brugada'] > 0).sum()}")
print(f"Healthy subjects: {(metadata['brugada'] == 0).sum()}")

Reading ECG Signal Data

The ECG recordings can be read using the wfdb package:

import wfdb
import matplotlib.pyplot as plt

# Read a single patient's ECG
patient_id = '188981'
record = wfdb.rdrecord(f'files/{patient_id}/{patient_id}')

# Access the signal data
signals = record.p_signal       # Signal array (samples × leads)
lead_names = record.sig_name    # Lead names (I, II, III, aVR, aVL, aVF, V1–V6)
sampling_freq = record.fs       # Sampling frequency (100 Hz)

# Plot a specific lead
plt.figure(figsize=(12, 4))
plt.plot(signals[:, 0])  # Plot first lead (Lead I)
plt.title(f'Patient {patient_id} - {lead_names[0]}')
plt.xlabel('Sample')
plt.ylabel('Amplitude (mV)')
plt.show()

This example reads the ECG waveform and visualizes a single lead for one patient. The same approach can be extended to batch processing, feature extraction, or model training.

Intended Uses

This dataset is intended to support:

• The development and validation of algorithms for the automatic detection and classification of Brugada syndrome.
• Comparative studies of ECG morphology between Brugada and normal patterns.
• Educational and research applications in computational cardiology.

Researchers can use the dataset to benchmark machine learning models, explore diagnostic features, or evaluate ECG interpretation methods in Brugada syndrome studies.

Limitations

While the dataset has been carefully curated, several limitations should be considered:

• The recordings represent a single-center cohort, which may limit generalizability to other populations.
• The dataset contains short-duration (12-second) recordings, which may not capture transient ECG changes.
• Diagnostic labels are binary (Normal vs. Brugada) and do not include detailed subtype classification (e.g., spontaneous vs. drug-induced).
• No preprocessing or filtering was applied, so users should apply appropriate signal processing techniques depending on their analysis goals.

Release Notes

Version 1.0.0: Initial public release of the dataset.

Ethics

All data were anonymized at the time of collection. Additionally, ethical approval for the data collection was obtained from the Regional Clinical Research Ethics Committee of the Principality of Asturias, No. 35/2013.

Conflicts of Interest

The authors declare no conflicts of interest.

References

1. Antzelevitch C, Brugada P, Borggrefe M, et al. Brugada syndrome: Report of the second consensus conference. Circulation. 2005;111(5):659–670.
2. Priori SG, Blomström-Lundqvist C, Mazzanti A, et al. 2015 ESC Guidelines for the management of patients with ventricular arrhythmias and the prevention of sudden cardiac death. Eur Heart J. 2015;36(41):2793–2867.
3. García-Iglesias D, de Cos FJ, Romero FJ, Polana S, Rubín JM, Pérez D, Reguero J, de la Hera JM, Avanzas P, Gómez J, et al. Spectral analysis of the QT interval increases the prediction accuracy of clinical variables in Brugada syndrome. J Clin Med. 2019;8(10):1629.
4. García-Iglesias D, Rubín J, Pérez D, Morís C, Calvo D. Insights for stratification of risk in Brugada syndrome. Eur Cardiol Rev. 2019;14(1):45.