RMSSD Estimation From Photoplethysmography and Accelerometer Signals Using a Deep Convolutional Network.

Area of Science:

• Biomedical Engineering
• Physiological Monitoring
• Machine Learning in Healthcare

Background:

• Heart Rate Variability (HRV) is a key indicator of Autonomic Neural System (ANS) function.
• Traditional HRV analysis relies on high-quality electrocardiogram (ECG) recordings.
• Photoplethysmography (PPG) sensors, common in wearables, offer a more accessible method for monitoring heart activity.

Purpose of the Study:

• To develop a deep learning model for estimating HRV metrics from PPG signals.
• To assess the feasibility of using lower-quality, cost-effective PPG sensors for HRV analysis.
• To evaluate the model's performance across diverse conditions.

Main Methods:

• A deep learning approach was designed to process PPG data.
• The model was trained to estimate the Root Mean Square of Successive Differences (RMSSD), a common HRV metric.
• The method was validated under various conditions using PPG sensor data.

Main Results:

• The deep learning model successfully estimated the RMSSD metric from PPG signals.
• Accurate HRV estimation was achieved even with lower-quality sensor data.
• The approach demonstrated robustness across a wide range of conditions.

Conclusions:

• Deep learning enables reliable HRV estimation from accessible PPG sensors.
• This method can potentially lower the cost and increase the availability of HRV monitoring.
• The findings support the use of consumer-grade wearables for physiological assessments.