A Heartbeat Classifier for Continuous Prediction Using a Wearable Device.

Area of Science:

• Biomedical Engineering
• Cardiovascular Health
• Wearable Technology

Background:

• Traditional heartbeat monitoring has limitations in early cardiovascular disease detection due to short recording times and lack of portability.
• Wearable devices like the Polar H10 offer continuous heartbeat recording but lack integrated data analysis capabilities.
• Automatic heartbeat classification remains a challenge for real-time health monitoring.

Purpose of the Study:

• To develop an automatic heartbeat classifier utilizing RR interval data for real-time, continuous monitoring.
• To evaluate machine learning and deep learning models for heartbeat classification using the Polar H10 device.
• To compare intra-patient and inter-patient training paradigms for optimal classification accuracy and computational speed.

Main Methods:

• A heartbeat classifier was developed using RR interval data from the Polar H10 wearable sensor.
• Various machine learning and deep learning algorithms were employed for classifier training.
• Comparative analysis of intra-patient and inter-patient paradigms was conducted on original and oversampled datasets.

Main Results:

• The random forest-based classifier, using RR interval data, achieved up to 99.67% in accuracy, precision, recall, and F1-score.
• Experiments were conducted to assess the classifier's performance in a real-time monitoring system with healthy individuals.

Conclusions:

• The proposed heartbeat classifier demonstrates high accuracy for real-time, continuous monitoring using wearable sensor data.
• This system holds significant potential for early cardiovascular disease detection and personalized health management.
• Further real-time evaluations are underway to validate the classifier's clinical applicability.