Authors
Hao Zhou, Md Mahbubur Rahman, Mehrab Bin Morshed, Yunzhi Li, Md Saiful Islam, Larry Zhang, Jungmok Bae, Christina Rosa, Wendy Berry Mendes, Jilong Kuang
Abstract
Cardiac Output (CO) is a critical indicator of health, offering insights into cardiac dysfunction, acute stress responses, and cognitive decline. Traditional CO monitoring methods, like impedance cardiography, are invasive and impractical for daily use, leading to a gap in continuous, non-invasive monitoring. Although recent advancements explored wearables on heart rate monitoring, these approaches face challenges in accurately estimating CO due to the indirect nature of the signals. To address these challenges, we introduce EarCO, a non-invasive multimodal CO monitoring system with Photoplethysmography and Ballistocardiogram signals on commodity earbuds. A novel feature fusion method is proposed to integrate raw signals and prior knowledge from both modalities, improving the system’s interpretability and accuracy. EarCO achieves an error of 1.080 L/min in the leave-one-subject-out settings with 62 subjects, making cardiovascular health monitoring accessible and practical for daily use.
Key Findings
- First multimodal earbud system: EarCO is the first work to utilize both PPG and BCG signals from commodity earbuds for cardiac output estimation
- Multimodal advantage: Combined PPG and BCG features improve estimation accuracy over single-modality approaches by over 23% in mean absolute error
- Feature fusion innovation: Novel feature fusion and guidance method based on first-order Taylor approximation enhances both accuracy and interpretability
- Clinical-grade performance: Achieved mean absolute error comparable to clinical devices despite using low-cost consumer earbuds
- Generalization capability: Leave-one-subject-out validation demonstrates the system’s ability to work across different users without individual calibration
- Complementary signals: PPG captures blood flow dynamics while BCG captures mechanical heart activities, providing comprehensive cardiovascular assessment
Methodology
The study used 62 healthy participants aged 18-51 with validated experimental protocols:
- Data Collection: Participants performed various stress tasks and relaxation activities while wearing commodity earbuds and reference BIOPAC equipment
- Signal Processing:
- PPG signals: Sampled at 25 Hz, bandpass filtered (0.7-3.5 Hz)
- BCG signals: Captured from built-in accelerometer at 50 Hz
- Feature Extraction:
- Deep features: Learned using Temporal Convolutional Networks (TCN) with causal convolution
- Surrogate biomarkers: Morphological features based on established physiological relationships
- Fusion Method: Novel feature fusion using first-order Taylor approximation to guide deep feature learning with surrogate biomarkers
- Validation: Leave-one-subject-out cross-validation to ensure generalizability
- Reference Standard: Impedance cardiography using BIOPAC NICO module for ground truth
Impact
This research demonstrates feasible cardiac output monitoring using consumer devices:
- Healthcare accessibility: Makes continuous cardiovascular monitoring available through everyday consumer devices
- Non-invasive monitoring: Provides alternative to invasive clinical methods for daily health tracking
- Preventive healthcare: Enables early detection of cardiovascular dysfunction and stress responses
- Wearable technology advancement: Establishes framework for multimodal physiological sensing in compact form factors
- Clinical translation: Bridges gap between clinical-grade monitoring and consumer health technology
- Daily life integration: Supports continuous cardiovascular assessment without disrupting normal activities
- Research foundation: Provides validated methodology for future development of earbud-based health monitoring systems