A pediatric ECG database with disease diagnosis covering 11643 children

Area of Science:

• Cardiology
• Medical Informatics
• Pediatrics

Background:

• Electrocardiogram (ECG) is vital for diagnosing cardiovascular diseases.
• Existing ECG datasets primarily focus on adults and lack disease-specific diagnoses.
• There is a significant need for comprehensive pediatric ECG data for AI-driven diagnostics.

Purpose of the Study:

• To develop and present a novel ECG database specifically for children aged 0-14 years.
• To include detailed cardiovascular disease diagnoses for the pediatric population.
• To facilitate the advancement of deep learning models for pediatric cardiovascular disease identification.

Main Methods:

• Collected 14,190 pediatric ECG records (12-lead and 9-lead) from 11,643 hospitalized children (2018-2024).
• Standardized ECG data with a 500 Hz sampling rate and 5-120 second record lengths.
• Encoded ECG records following AHA/ACC/HRS guidelines and Chinese expert consensus, assigning labels for 19 common pediatric cardiovascular diseases.

Main Results:

• The dataset contains 14,190 pediatric ECG records.
• 3,516 records are diagnosed with cardiovascular diseases, covering 19 common conditions.
• Includes conditions such as myocarditis, cardiomyopathy, congenital heart disease, and Kawasaki disease.

Conclusions:

• This curated pediatric ECG database addresses a critical gap in existing resources.
• The dataset provides valuable, diagnosed data for training and validating deep learning algorithms in pediatric cardiology.
• Enables improved intelligent diagnosis of cardiovascular diseases in children.