Abstract
Cardiovascular diseases claim over 10 million lives annually, highlighting the critical need for long-term monitoring and early detection of cardiac abnormalities. Existing techniques like electrocardiograms (ECG) and Holter are accurate but suffer from discomfort caused by body-attached electrodes. While wearable devices using photoplethysmography offer more convenience, they sacrifice accuracy and are susceptible to environmental interference. Here we present a radio frequency (RF)-based sensing system that monitors long-term heart rate variability (HRV) with clinical-grade accuracy. For the first time, our system successfully overcomes the orders-larger interference from respiration motion by overturning the conventional signal processing framework. By identifying previously undiscovered frequency ranges where heartbeat information predominates over other motions, we generate prominent heartbeat patterns with harmonics typically considered detrimental. Extensive evaluations, including a large-scale clinical setting involving 7,090 participants and a long-term daily scenario spanning two months, demonstrate that our system can monitor HRV and diagnose abnormalities with comparable performance to clinical-grade ECG-based systems. This RF-based HRV sensing system has the potential to support active self-assessment and revolutionize medical prevention with long-term and precise health monitoring.
The Challenge of Continuous Monitoring
Continuous cardiac monitoring is essential for the early detection of cardiovascular diseases (CVDs), yet current tools force a trade-off between comfort and accuracy. Clinical standards like ECG and Holter monitors require body-attached electrodes that can irritate the skin and discourage long-term use. Conversely, wearables like smartwatches using photoplethysmography (PPG) are more convenient but struggle with skin tone variations, ambient light, and motion artifacts.
Contactless radio frequency (RF) sensing, or radar, has long been considered the "holy grail" of unobtrusive monitoring. However, it faces a fundamental physics challenge: the chest displacement from breathing (centimeter-scale) is orders of magnitude larger than the displacement from a heartbeat (sub-millimeter scale). In the frequency domain, "spectral leakage" from respiration usually overwhelms the heartbeat signal, making precise heart rate variability (HRV)—the millisecond-level timing between beats—notoriously difficult to capture.
Methodology: Turning Noise into Signal
The researchers developed RF-HRV, a system that "overturns" the conventional signal processing framework. Instead of attempting to filter out respiration to find the fundamental heartbeat frequency, the system looks higher up the spectrum.
Key Technical Innovations:
- Signal Selection: The system uses a voxel-based selection algorithm to identify the specific reflection point on the torso with the richest heartbeat information.
- Harmonic Leveraging: Traditionally, higher-order harmonics are viewed as detrimental interference. However, this paper demonstrates that respiratory harmonics decay much faster than heartbeat harmonics. By using Variational Mode Decomposition (VMD), the system extracts high-frequency components where the heartbeat signal finally dominates the respiratory "noise."
- The Beat Frequency Effect: To reconstruct a clean timing signal, the system superimposes adjacent heartbeat harmonics. This creates a "beat frequency" signal that matches the actual heart rate but exists in a much cleaner, high-frequency environment.
- Dynamic Peak Seeking: A dynamic thresholding algorithm identifies individual heartbeat peaks to calculate the Intra-Beat Interval (IBI) with millisecond precision.
Large-Scale Clinical Validation
The study is remarkable for its scale, moving beyond small lab groups to real-world clinical environments. The system was validated against professional 12-lead ECGs and Polysomnography (PSG) equipment across two primary datasets:
- Large-scale Clinical Dataset: 7,090 participants at a tertiary hospital, ranging from ages 10 to 94, covering various cardiac conditions.
- Long-term Daily Scenario: A longitudinal study of a single individual over 5 nights of sleep, spanning a two-month period.
Performance Results
In the clinical setting, RF-HRV achieved a median Real-Time IBI (RT-IBI) error of 26.1 ms, representing a 10x accuracy improvement over previous state-of-the-art wireless methods. Using only the wireless IBI data, the system achieved an average classification accuracy of 83.4% for identifying cardiac conditions, with particularly high effectiveness in diagnosing Bradycardia (92.2%) and Tachycardia (93.9%).
Discussion and Future Directions
The primary insight of this research is that harmonics are friends, not enemies. By proving that clinical-grade wireless sensing is possible across a massive age demographic (10–94 years), the authors have shown that RF-HRV is viable for general population health screening.
While the system excels at rhythm-based detection, it currently struggles with complex Arrhythmia (47.6% accuracy) due to simplified classification rules. Future work will focus on integrating machine learning models to improve the detection of irregular rhythms and enhancing signal selection to remain robust against "tossing and turning" during sleep. This technology represents a significant step toward a future where precise health monitoring is integrated seamlessly into our daily environments.
Summary updated