Proceedings of the 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference Shanghai, China, September 1-4, 2005
Noninvasive Estimation of Blood Pressure Using Photoplethysmographic Signals in the Period Domain Y.S.Yan and Y.T.Zhang Joint Research Center for Biomedical Engineering, Department of Electronic Engineering, The Chinese University of Hong Kong, Hong Kong
[email protected] Abstract—In order to detect hypertension at its early stage, blood pressure (BP) of an individual must be carefully monitored. Aiming to monitor BP noninvasively and continuously, this study examines a new feature, normalized harmonic area (NHA), which is extracted from photoplethysmographic (PPG) signals in the period domain by using the discrete period transform (DPT). BP, pulse transit time (PTT), diastolic time (DT) and NHA were obtained from 28 subjects before and immediately after step-climbing exercise. It was found that NHA has more significant correlation with BP than PTT and DT in this study. The mean difference and standard deviation (Mean f SD) between the BP estimated from NHA and the reference BP obtained from a commercial oscillometric BP meter were 0.37f4.3 mmHg and 0.51f4.8 mmHg for SBP and DBP, respectively. The result indicate that NHA is potentially a useful indicator of arterial BP.
18 seconds at sampling rate of 1 KHz with the DATAQ instrument WINDAQ. The blood pressure was measured by an automatic oscillometric BP meter (Model BP-8800, COLIN) at left upper arm immediately after recording the ECG and PPG. The experiments consisted of three sessions for different blood pressure measurements: at rest, immediately after step-climbing exercise, and during recovery. For each subject, 3 trials were recorded in each of the 3 sessions. The typical waveforms of subject A during different sessions with blood pressure changes were shown in Fig.1.
I. INTRODUCTION Hypertension is an important risk factor for cardiovascular diseases, such as cardiac failure, stroke and coronary artery disease. In order to detect hypertension at its early stage, frequent and regular checking of an individual’s blood pressure is very important. Most of the current noninvasive methods for BP measurement are not suitable for continuous and long term monitoring because they commonly use an occlusive cuff. The use of the cuff has several disadvantages, such as underestimation of systolic BP, false readings caused by improper cuff size, and patient discomfort, etc. In previous studies, pulse transit time (PTT) was reported as an accurate noninvasive and non-occlusive technique of BP measurement [1,2]. However, PTT techniques require two independent channels to measure the time interval, which are less convenient to users in practical application. For this reason, some attempts were made to estimate BP by using photoplethysmographic (PPG) signal. Several features (pulse amplitude, area, width, half width, systolic and diastolic time (DT), etc.) extracted from PPG signals in time domain were claimed to correlate with BP [3,4]. In this study, a new feature of PPG in the period domain instead of in the time domain is investigated. II. METHODOLOGY Twenty-eight healthy volunteers, aged 24-30 yrs., participated in the experiment. The ECG signals were recorded by standard lead I, and PPG signals were collected simultaneously by a self-designed reflectance probe at the right hand index finger tip. Both signals were recorded for 0-7803-8740-6/05/$20.00 ©2005 IEEE.
Fig.1. Typical PPG recordings from subject A: (a) at rest, (b) immediately after exercise, and (c) during recovery.
As shown in Fig.1, the position and depth of dicrotic notch changed with different test conditions. This is because exercise would cause the changes of cardiac output, and peripheral resistance, as well as BP [5]. In other words, the changes in the position and depth of the dicrotic notch might indicate changes in BP. To quantify the changes of position and depth of dicrotic notch, we propose to evaluate the changes of distribution of harmonic components of PPG waveform. We separated the whole PPG pulse train into individual pulses according to local maxima (or minima), and calculated the spectrum of each beat in period domain by using the discrete period transform (DPT) algorithm [6,7]. This algorithm has the advantage of improved resolution for low frequency signals such as PPG signals (approx. 0.1~10 Hz), compared with the conventional FFT algorithm. As illustrated in Fig.2, the normalized harmonic area (NHA) is defined as the normalized shadow area:
3583
NHA
k2
¦ DPT (tn )
n 0
k1
¦ DPT (tn ) , n 1,2, N ,
n 0
(1)
where tk1 and tk2 is the period (or inverse of frequency) at the 1st and 2nd harmonic of the DPT spectrum, and DPT(t) is the amplitude of the spectrum component at the period t.
of noninvasive arterial pressure, obtained from at least 85 patients, should not exceed 5±8 mmHg from a reference method [4]. TABLE II Comparison of Mean Differences and Standard Deviations of BP Estimation with Calibration Using the Examined Features PTT SBP* DBP*
0.43f6.3 0.47f5.5 * MeanfSD (mmHg)
DT
NHA
0.60f7.9 0.48f5.2
0.37f4.3 0.47f4.8
Fig.2.Illustration of DPT spectrum: P1 and P2 label the peaks of the 1st and 2nd harmonic component of spectrum, respectively. NHA is defined as the normalized shadow area.
IV. CONCLUSION
For each subject, correlation coefficients between BP and NHA were calculated from 9 trials of three experiment sessions. For BP estimation, the data obtained from three trials were used to set up calibration curve y=a BP+ b by linear regression and the data from the other 6 trials were used to estimate BP for each subject. As a comparison, PTT and DT were also examined for BP estimation. In this study, PTT was estimated as the time interval between ECG R wave and peak of PPG signal, and DT was detected with wavelet transform as described in [4].
In this paper, NHA, a new feature extracted from photoplethysmographic signals in the period domain, is proposed to measure BP noninvasively and continuously. The experimental results show that NHA has a higher correlation with BP and smaller BP estimation error than both PTT and DT, which indicates that NHA is potentially a useful indicator of arterial BP. Future research will attempt to explore the physiological mechanisms behind the close relation between NHA and BP, and validate the proposed approach with more subjects under various conditions.
III. RESULTS
ACKNOWLEDGMENT
Table I shows the intra-subject correlation of SBP and DBP with PTT, DT and NHA. For NHA, almost all of the correlations with SBP (28 of 28) and DBP (27 of 28) were negative. Most of the correlations (25 for SBP, 23 for DBP) were significant (p
