JoWUA

Abstract

Hypertension is the predominant cardiovascular disease (CVD) risk factor, but the standard method of blood pressure (BP) assessment has proven unreliable. This has led to an increasing interest in developing more sophisticated and accurate methods. In this paper, we present a novel technique that incorporates deep learning, sensor fusion, and feature selection strategies for precise blood pressure measurement. This approach requires no additional records beyond the data from blood pressure cuffs, ECG, photoplethysmography (PPG), and sensor data, and features with a strong one-to-one positive relationship with blood pressure are identified. A deep-learning model using these features, sensor fusion, and feature selection techniques is trained over the dataset. Deep learning model - trained on a large dataset of blood pressure values to know the complex correlations between the features and blood pressure. The results of our research, presented in Section 6, show that our novel methodology significantly increased the accuracy of the blood pressure rate by 8% on average. Our deep learning approach is orders of magnitude more sensitive and substantially more resistant to sensor noise and artifacts compared to classical approaches, allowing even minute changes in blood pressure measurements to be accurately captured. This makes our method applicable in practice where accurate and reliable blood pressure measurements are required. Crucially, we can also apply our method to blood pressure monitoring in chronic patients for telemedicine. In this sense, considering the increasing omnipresence and acceptance of wearable devices, we foresee the integration of our proposed methodology, hopefully in the form of such wearables, to enable noninvasive and continuous blood pressure (BP) monitoring. In general, this approach offers a promising alternative for improving the accuracy of BP rate measurements. This work demonstrates the potential to leverage information from multiple sensors by utilizing deep learning for sensor fusion and feature selection processes, resulting in more accurate and reliable blood pressure (BP) measurements. That could make a significant difference in the accuracy of diagnosing or treating high blood pressure with the potential to improve patient outcomes and reduce healthcare costs in the long run.