Efficient Blood Pressure Estimation Using Seismocardiogram and Electrocadiogram

Our research team has developed a device for monitoring cardiac signals in the carotid artery. This device enables the simultaneous measurement of various cardiovascular signals, including Photoplethysmogram (PPG), Phonocardiogram (PCG), Electrocardiogram (ECG), and Seismocardiogram (SCG). This study estimates blood pressure based on cardiac signals measured in the carotid artery and utilizes deep learning to minimize the sensors requirements for the estimation. Our approach focused on utilizing Seismocardiogram (SCG) and Electrocardiogram (ECG) signals for blood pressure estimation. By strategically leveraging the complementary nature of these signals, we aim to streamline the estimation process, reducing the need for extensive sensor fusion. To this end, our teams adapted the Long Short-Term Memory (LSTM) architecturefor for minimalist sensor fusion model. This model is trained on the Chemical Effects in Biological Systems (CEBS) database to extract features from sensor data. Additionally, the model demonstrates the ability to restore ECG data from SCG inputs, highlighting the feasibility of predicting ECG signals using only SCG sensors. This research contributes a resource-efficient paradigm to blood pressure estimation, prioritizing simplicity and minimizing sensor requirements. By emphasizing the independent utilization of SCG sensors in blood pressure estimation, our approach provides a practical solution for developing wearable health monitoring devices with reduced complexity and enhanced efficiency.