Development of a long term dynamic blood pressure monitoring system using cuff-less method and pulse transit time

Abstract

Hypertension has been a silent killer that has led to the death of millions of people annually. In this study, we developed a long term cuff-less blood pressure monitoring system that uses an office armchair with integrated sensors and communications circuits. The system is able to non-invasively acquire electrocardiography (ECG), photoplethysmography (PPG), and ballistocardiography (BCG) signals from the sensors placed on the gasbags that are in contact both the left and right legs upon detection of a person sitting down. These signals are acquired and pre-processed in a MCU, which transmits the data to a backend PC through a Bluetooth connection. The backend PC performs the blood pressure estimation using the ECG, PPG, and BCG waveforms processed using digital signal filtering, signature detection, time parameter detection, and blood pressure estimation processes. A total of 10 volunteers participated in the experiments wherein results from a commercial blood pressure monitoring device is compared with the proposed system that extracts BCG J peak from the left leg (RJI_L) and right leg (RJI_R) through sensors mounted on gasbags. RJI_L and RJI_R manifested a high collinearity and a variance inflation factor of up to 10 from the interval between the ECG R peak and the BCG J peak (RJI). Accordingly, RJI_R or RJI_L could be treated as independent variables for multiple regression analysis because they had similar correlation properties. An absolute mean pulse arrival time (PAT) of 0.910 and 0.659 was obtained for systolic blood pressure (SBP) and diastolic blood pressure (DBP), respectively. Furthermore, a moderate correlation was found between RRI and DBP while the RRI and SBP did not exhibit any linear correlation. The correlation coefficient r ranged between 0.9 and 1 in simple and multiple regression schemes. The correlation coefficient r improved to 0.757 when DBP was estimated in three- and four-variable regressions with PAT, RRI, and RJI_R as the independent variables. And a correlation coefficient of 0.774 was found between the proposed method and a traditional cuff approach for both SBP and DBP estimations. Therefore, the cuff-less method provides not only convenience to the user but also demonstrated the efficacy of the proposed method using pulse transit time.