Autoencoder-Based Extrasystole Detection and Modification of RRI Data for Precise Heart Rate Variability Analysis

Area of Science:

• Cardiovascular physiology
• Biomedical engineering
• Machine learning in healthcare

Background:

• Heart rate variability (HRV) analysis is crucial for autonomous health evaluation.
• Arrhythmias like premature ventricular contractions (PVC) and premature atrial contractions (PAC) distort HRV metrics.
• Accurate HRV analysis requires appropriate modification of ectopic R-R intervals (RRIs).

Purpose of the Study:

• To propose a unified framework for detecting and modifying ectopic RRIs caused by PVC and PAC.
• To enhance the accuracy of heart rate variability analysis in the presence of common arrhythmias.
• To develop a real-time system for reliable health monitoring.

Main Methods:

• Utilized an autoencoder (AE) for real-time ectopic RRI detection (AED).
• Employed a denoising autoencoder (DAE) to modify detected ectopic RRIs (DAEM).
• Applied the framework to real-world RRI data containing PVC and PAC.

Main Results:

• AED achieved 93% sensitivity and a low false positive rate (0.08/hour).
• DAEM significantly reduced root mean squared error: 31% for PVC, 73% for PAC.
• The framework successfully suppressed false positives in an epileptic seizure dataset.

Conclusions:

• The proposed AE/DAE framework offers accurate detection and modification of ectopic RRIs.
• This system can significantly improve the reliability of HRV-based health monitoring.
• The framework shows potential for advanced medical sensing systems.