Monitoring long-term cardiac activity with contactless radio frequency signals

A research team led by Prof. Chen Yan at the University of Science and Technology of China (USTC) has taken a leap forward in cardiovascular health monitoring. They have developed a non-invasive radio frequency (RF)-based system capable of monitoring heart rate variability (HRV) with clinical-grade accuracy over extended periods.

This research is published in Nature Communications.

Cardiovascular diseases (CVDs) are the leading cause of death globally, claiming around 17.9 million lives annually. In China, an aging population has further heightened the prevalence and mortality rates of CVDs. Research shows that early diagnosis and intervention can effectively prevent many cases of cardiovascular illness.

However, existing detection techniques like electrocardiogram (ECG) and Holter are accurate but have drawbacks. The electrodes attached to the body in ECG and Holter can cause discomfort, making them unsuitable for long-term use. Wearable devices, while more convenient, are less accurate and vulnerable to environmental interference.

The RF–HRV system developed by the research team successfully overcomes the interference from respiratory motion in far-field conditions by analyzing RF signals. The system employs a signal selection algorithm to identify the signal rich in heartbeat information from multiple reflected signals. It uses the variational mode decomposition (VMD) algorithm to extract high-frequency components, obtaining clear and accurate heartbeat patterns.

By superimposing adjacent heartbeat harmonics, it generates specific heartbeat patterns with a frequency equal to the heart rate to calculate HRV.

Additionally, the researchers evaluated the system in a large-scale outpatient setting (with 6,222 participants) and a long-term daily life scenario (continuous multi-night sleep monitoring). The results show that in the outpatient scenario, the median real-time inter-beat interval (RT-IBI) error of the system is 26.1 milliseconds, and in the daily scenario, it is 34.1 milliseconds, which is a significant improvement compared to existing systems that extract signals only from the heart rate frequency band.

Moreover, the system performs well in the automatic classification of heartbeat abnormalities and is comparable to clinical-grade 12-lead ECG systems.

The innovation of this study lies in breaking the traditional signal processing framework. It utilizes previously unexplored high-frequency ranges (beyond 10-order heartbeat harmonics) to extract heartbeat signals, overcoming the interference of respiratory motion on heart rate monitoring.