Classifier-Guided Diffusion Model for Generating Disease-Specific ECG Data: A Case Pilot Study on Inferior Myocardial...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Electrocardiography (ECG) is a crucial non-invasive cardiac screening tool.
• High-quality, annotated ECG data is scarce due to privacy and resource limitations.
• Existing synthetic ECG generation methods require costly labeled datasets.

Purpose of the Study:

• To develop a cost-effective method for generating disease-specific synthetic ECG data.
• To reduce reliance on extensive annotations for synthetic ECG generation.
• To generate inferior myocardial infarction (IMI) ECG data using a novel approach.

Main Methods:

• Employed a classifier-guided diffusion model for synthetic ECG generation.
• Applied the method to the PTB-XL dataset to generate IMI ECGs.
• Evaluated synthetic data quality through classification performance and statistical analysis.

Main Results:

• Classifiers trained on merged real and synthetic IMI ECG data performed comparably to those trained solely on real data.
• Statistical analysis showed generated IMI ECGs closely matched real IMI ECGs in Q wave depth distribution.
• The proposed method effectively generates high-quality, reliable synthetic ECG data.

Conclusions:

• Synthetic ECG data can partially replace real data without compromising diagnostic performance.
• The classifier-guided diffusion model offers a viable solution for data scarcity in cardiac research.
• This approach validates the quality and reliability of generated synthetic ECG data for IMI.