A 12-lead electrocardiogram database for arrhythmia research covering more than 10,000 patients

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Arrhythmias like atrial fibrillation significantly impact public health and healthcare costs.
• Long-term electrocardiogram (ECG) monitoring is crucial for diagnosing cardiac conditions but generates vast, complex datasets.
• Analyzing large ECG datasets requires significant expert time and effort.

Purpose of the Study:

• To introduce a novel, comprehensive database of 12-lead ECG signals for research.
• To facilitate the development and validation of machine learning and statistical methods for cardiovascular condition analysis.
• To support the scientific community in advancing the study of arrhythmias and other cardiac diseases.

Main Methods:

• Collected and curated a dataset of 12-lead ECGs from 10,646 patients.
• Recorded ECGs at a 500 Hz sampling rate, with each recording lasting 10 seconds.
• Included expert-labeled data covering 11 common rhythms and 67 additional cardiovascular conditions.

Main Results:

• The database contains 10-second, 12-dimension ECG recordings.
• Each ECG is meticulously labeled with rhythm and condition information by professional experts.
• The dataset is structured to enable comparative analysis of various machine learning and statistical techniques.

Conclusions:

• The newly established ECG database provides a valuable resource for cardiovascular research.
• This dataset will accelerate the development of automated diagnostic tools for arrhythmias and other heart conditions.
• It enables researchers to design, compare, and refine novel and classical analytical methods.