Cuff-Less Blood Pressure Estimation Using Pulse Waveform Analysis and Pulse Arrival Time

• Published in: IEEE journal of biomedical and health informatics Vol. 22; no. 4; pp. 1068 - 1074
• Main Authors: Yoon, Young-Zoon, Kang, Jae Min, Kwon, Yongjoo, Park, Sangyun, Noh, Seungwoo, Kim, Younho, Park, Jongae, Hwang, Sung Woo
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.07.2018; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Biomedical monitoring; Blood pressure; Calibration; Correlation; Cuff-less blood pressure measurement; Data models; Electrocardiography; Estimation; Indexes; Model accuracy; photoplethysmography; Pressure measurement; pulse arrival time; pulse waveform analysis; Regression analysis; Regression models; Scaling; Wave analysis; wearable device
• ISSN: 2168-2194; 2168-2208; 2168-2208
• DOI: 10.1109/JBHI.2017.2714674

Using the massive MIMIC physiological database, we tried to validate pulse wave analysis (PWA) based on multiparameters model whether it can continuously estimate blood pressure (BP) values on single site of one hand. In addition, to consider the limitation of insufficient data acquirement for home user, we used pulse arrival time (PAT) driven BP information to determine the individual scale factors of the PWA-BP estimation model. Experimental results indicate that the accuracy of the average regression model has error standard deviations of <inline-formula> <tex-math notation="LaTeX">10.6\pm 3.3</tex-math></inline-formula> mmHg (PAT), <inline-formula> <tex-math notation="LaTeX">8.7\pm 3.2</tex-math></inline-formula> mmHg (PWA) for SBP and <inline-formula> <tex-math notation="LaTeX">6.0\pm 2.3</tex-math></inline-formula> mmHg (PAT), <inline-formula> <tex-math notation="LaTeX">4.4\pm 1.6</tex-math></inline-formula> mmHg (PWA) for DBP on 23 subjects over a 1 day period. We defined a local-model which is extracted regression model from sparsely selected small dataset, contrast to full dataset for 24h (average-model). The limit of BP estimation accuracy from the local-model of PWA is lower than that of PAT-BP average-model. Whereas the error of the BP estimation local-model was reduced using more data for scaling, it required more than four times the 1 min data extracted over the 12 h calibration period to predict BP for 1 day. This study shows that PWA has possibility to estimate BP value and PAT-driven BP information could be used to determine the individual scale factors of the PWA-BP estimation model for home users.