Prediction of Sudden Cardiac Death Risk with a Support Vector Machine Based on Heart Rate Variability and Heartprint...

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Sudden cardiac death (SCD) prediction typically relies on lengthy electrocardiogram (ECG) recordings.
• Existing methods utilize heart rate variability (HRV) or heartprint indices over 24-hour periods.
• Short-term ECG analysis for SCD prediction remains an area for advancement.

Purpose of the Study:

• To identify optimal combinations of HRV and heartprint indices for predicting SCD using short-term ECG recordings (1000 heartbeats).
• To develop and validate a Support Vector Machine (SVM) model for early SCD risk assessment.

Main Methods:

• Measured eleven HRV indices and five heartprint indices from 135 ECG recording pairs (pre-SCD vs. control).
• Employed Support Vector Machines with a radial basis function kernel and hyperparameter optimization.
• Utilized 10-fold cross-validation to systematically identify the best index combinations and optimize SVM parameters (gamma and cost).

Main Results:

• The best index combinations achieved high predictive performance, with Area Under the Curve (AUC) values ranging from 0.80 to 0.86 and accuracies from 80% to 86%.
• Validation on an independent dataset yielded an AUC of 0.68 and accuracy of 67% for the top-performing combination.
• The study systematically identified 13 superior combinations of HRV and heartprint indices.

Conclusions:

• Short-term ECG recordings, analyzed with optimized SVM models, can effectively predict SCD risk.
• The developed SVM model shows potential for early warning systems, aiding in the prevention of imminent SCD.
• This approach offers a more efficient alternative to long-term monitoring for SCD risk assessment.