Novel Continuous Monitoring of Cardiac Health

University of Pittsburgh researchers have developed a novel approach to continuously monitor blood pressure (BP). Using a diffusion-based generative model, it is possible to convert photoplethysmography (PPG) signals into electrocardiogram (ECG) quality outputs to accurately continuously monitor BP. This novel approach can also generate an ECG signal with ECG characteristics (e.g., QRS complex data) from PPG signals, removing the need for complex, multi-lead ECG analysis. Development of this technology could dramatically improve the accessibility of continuous BP monitoring for early detection of cardiovascular diseases (CVDs).

Description

High BP, hypertension, is a major risk factor for CVDs. Accurate continuous monitoring of BP during daily living (i.e., not in a clinical setting) is a key tool in early detection and treatment of CVDs. Early recognition of hypertension can also facilitate interventions to prevent CVDs from developing. There is a need to develop accurate, cost-effective methods to continuously monitor BP. PPG can be used to monitor BP using wrist worn wearable devices but lacks the accuracy of the more complex ECG method. This unique approach can provide ECG-level accuracy and performance using only a PPG signal. Widespread use of this technology could improve the accessibility and accuracy of continuous BP monitoring to clinicians, leading to early detection, prevention, or treatment of CVDs in patients.

Applications

• Continuous blood pressure monitoring

Advantages

The gold standard for continuous BP monitoring is ECG. However, this is costly and an unpleasant experience for patients, requiring attachment of straps and patches to the patient’s chest. Given the accessibility of PPG-based technologies for continuous BP monitoring, efforts to use PPG data to generate the more accurate ECG signals have been previously attempted. To date, these approaches fail to accurately model ECG characteristics, meaning a loss of critical details of cardiac function including R-peaks, signal amplitude range, and QRS complex.

This novel approach is designed to accurately generate ECG signals from user-friendly PPG-based technologies. The end-to-end PPG-conditional generative model-based framework (PPGG) includes a QRS adaptive search-guided forward process to locate the QRS complex, pivotal in BP determination, and selectively adds noise to key regions for more targeted generation of ECG signals with improved quality. The PPGG also includes a reverse process to enhance the generated ECG signal further. A bidirectional long short-term memory model uses both the generated ECG and PPG input to estimate blood pressure.

Invention Readiness

The PPGG model was trained and validated using two publicly available datasets containing PPG and ECG data. This PPGG model could generate ECGs with high fidelity to the original ECGs. Real-world testing confirmed the feasibility of using PPG data and the PPGG model to continuously monitor BP. The PPGG model could successfully run on both a laptop and a smartphone suggesting this approach could be used for real-time BP estimation and ECG generation using commercially available technology, vital for improving accessibility to continuous BP measurements.

IP Status

Patent Pending

Related Publication(s)

Ji, H., & Zhou, P. (2024). Advancing PPG-Based Continuous Blood Pressure Monitoring from a Generative Perspective. Proceedings of the 22nd ACM Conference on Embedded Networked Sensor Systems, 661–674. https://doi.org/10.1145/3666025.3699365