Key result
Deep regression model accurately estimates continuous blood pressure with an SBP error of ~3.6 mmHg.
Why the study?
Existing cuffless blood pressure monitoring methods using PPG and ECG signals suffer from instability and susceptibility to various factors, hindering accurate estimation.
Does a deep regression model with state space reconstruction improve the accuracy of continuous blood pressure estimation from PPG and ECG signals?
Does a deep regression model with state space reconstruction improve the accuracy of continuous blood pressure estimation from PPG and ECG signals?
A novel deep regression model with state space reconstruction provides highly accurate continuous blood pressure estimation from PPG and ECG signals, meeting British Hypertension Society standards.
Answers are grounded in this paper and cite their sources.
Supports cuffless continuous BP monitoring development; leaves open prospective clinical validation before adoption.
Continuous blood pressure (BP) monitoring is crucial for diagnosing and preventing cardiovascular disease (CVD). However, existing approaches for continuous cuffless BP monitoring using photoplethysmogram (PPG) and electrocardiogram (ECG) signals suffer from instability and susceptibility to various factors. This poses a challenge in accurately estimating BP levels, hindering effective disease diagnosis and prevention. Therefore, there is a need for an improved method that can provide reliable and accurate continuous BP estimation from PPG and ECG signals, overcoming the limitations of existing approaches. In this study, we proposed a deep regression model with state space reconstruction (SSR) for continuous BP estimation. A feature voting system with a variety of feature selection algorithms is introduced to select the optimal feature set of PPG and ECG signals. The SSR technique is applied to feature data to reveal useful hidden information. The proposed method is evaluated based on 660 subjects from a well-known benchmark dataset and a multi-day BP dataset. Random forest and the proposed deep regression model are tested to show the advantages of using SSR on feature data. The results show a promising performance of the improved deep regression model. On the benchmark dataset, the root mean square error (RMSE) and mean absolute error (MAE) of the improved deep regression model are 3.613 and 2.765 mmHg respectively for systolic BP (SBP), 1.978 and 1.543 mmHg for diastolic BP (DBP). The results achieved high accuracy for estimating SBP and DBP according to the British Hypertension Society (BHS) standard. On the multi-day BP dataset, the proposed model achieved RMSE of 5.387, 3.338 and 3.611 mmHg for SBP and achieved MAE of 4.115, 2.553, and 2.927 mmHg for DBP. Additionally, the robustness of our proposed model is validated by adding random noise into the PPG signals. The results demonstrate that the proposed deep regression model with SSR can improve the performance of BP estimation. It is possible to apply our proposed method further to develop a wearable device for real-time BP monitoring.
No takes yet. Share an insight, caveat, or question.
Lyu et al. (2024) studied Continuous blood pressure estimation (n=660). Deep regression model with state space reconstruction (SSR) vs. Random forest was evaluated on Root mean square error (RMSE) and mean absolute error (MAE) for systolic and diastolic blood pressure. A deep regression model with state space reconstruction accurately estimated continuous blood pressure, achieving an RMSE of 3.613 mmHg for SBP and 1.978 mmHg for DBP on a benchmark dataset.
Synapse has enriched 10 related papers. Consider them for comparative context: