Non-linear dynamics in ECG: a novel approach for robust classification of cardiovascular disorders

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Electrocardiogram (ECG) analysis is crucial for cardiac care.
• Current ECG methods struggle with waveform variability, signal non-linearity, and low amplitudes.
• Non-linear dynamics offer potential for improved ECG interpretation.

Purpose of the Study:

• To develop and validate a non-linear analysis approach for detecting cardiac disorders from multi-channel ECG.
• To leverage Recurrence Plots for analyzing complex ECG patterns.
• To assess the classification accuracy of the proposed method across various cardiac conditions.

Main Methods:

• Utilized the Physikalisch-Technische Bundesanstalt (PTB) dataset from PhysioNet.
• Applied Recurrence Plot visualizations to ECG signals to capture patterned occurrences.
• Employed Recurrence Quantitative Analysis (RQA) on extracted features.
• Performed Wilcoxon rank-sum tests for statistical significance.
• Used t-SNE for visualizing latent space embeddings.

Main Results:

• Achieved 100% classification accuracy for detecting cardiac disorders (Myocardial infarction, Bundle branch blocks, Cardiomyopathy, Dysrhythmia) and healthy controls.
• Identified five statistically significant RQA features differentiating study groups at 95% confidence interval.
• t-SNE visualizations demonstrated clear separation between cardiac disorders and healthy subjects.

Conclusions:

• The proposed non-linear analysis using Recurrence Plots is highly effective for cardiac disorder detection.
• Recurrence Plots and RQA features provide robust biomarkers for differentiating cardiac conditions.
• This approach offers a promising advancement in automated ECG interpretation for improved cardiac diagnostics.