Wrist-Worn Heartbeat Monitoring System Based on Bio ... - IEEE Xplore

23 downloads 1793 Views 929KB Size Report
based wrist-worn heartbeat monitoring system is proposed. The system is able to estimate heart rate from a subject's wrist with only four electrodes. The design ...
Wrist-worn Heartbeat Monitoring System Based on Bio-Impedance Analysis* Jia Xu, Xiaomeng Gao, Student Member, IEEE, Alexander Lee, Shuhei Yamada, Ehsan Yavari, Student Member, IEEE, Victor Lubecke, Fellow, IEEE, and Olga Boric-Lubecke, Fellow, IEEE 

Abstract—In this paper, a bio-impedance analysis (BIA) based wrist-worn heartbeat monitoring system is proposed. The system is able to estimate heart rate from a subject’s wrist with only four electrodes. The design is achieved with a standard BIA device and off-the-shelf components for signal conditioning. The measured heartbeat-related impedance signal is compared with a reference heart rate signal obtained from piezoelectric finger pulse transducer. The BIA results agree with the reference, which validates the feasibility of the proposed system. To the best of our knowledge, this is the first reported BIA heartbeat monitoring system in the wristband configuration.

I. INTRODUCTION Heart disease (which includes Heart Disease, Stroke and other Cardiovascular Diseases) is the No. 1 cause of death in the United States. People in all ages and backgrounds can get the condition. According to the American Heart Association, about 650,000 people were killed by heart disease in 2011 [1]. Early detection of heart disease enables timely and more effective treatments, thus opening up the important possibilities of preventing delayed disease diagnosis or even fatal consequences. The heartbeat signal is one of the most important indicators of heart problems. Monitoring heartbeat signals during everyday life is not only helpful for early detection of heart disease, but also helps monitor health status of the subjects afflicted with such conditions and patients in rehabilitation state. The rising need for everyday health care at home also makes heartbeat monitoring devices necessary. Currently Electrocardiogram (ECG) and finger pulse *Research supported by Archinoetics LLC. The views, opinions and/or findings contained in this paper are those of the authors and should not be construed as an official Department of the Army position, policy or decision unless so designated by other documentation. Jia Xu is with the Electrical Engineering Department, University of Hawaii at Manoa, Honolulu, HI 96848 USA (e-mail: [email protected]). Xiaomeng Gao is with the Electrical Engineering Department, University of Hawaii at Manoa, Honolulu, HI 96848 USA (e-mail: [email protected]). Alexander Lee is with the Electrical Engineering Department, University of Hawaii at Manoa, Honolulu, HI 96848 USA (e-mail: [email protected]). Shuhei Yamada is with the Department of Electrical Engineering, University of Hawaii at Manoa, Honolulu, HI 96848 USA (e-mail: [email protected]). Ehsan Yavari was with the Department of Electrical Engineering, University of Hawaii at Manoa, Honolulu, HI 96848 USA (e-mail: [email protected]). Victor Lubecke is with the Electrical Engineering Department, University of Hawaii at Manoa, HI 96822 USA, (e-mail: [email protected]). Olga Boric-Lubecke is with the Electrical Engineering Department, University of Hawaii at Manoa, HI 96822 USA, (e-mail: [email protected]).

978-1-4577-0220-4/16/$31.00 ©2016 IEEE

transducer are two commonly used devices in hospital settings for continuous heartbeat monitoring. Although ECG has high accuracy that provides detailed information about cardiac activity, the requirement of professional nursing personnel for electrodes configuration makes it difficult to practice outside the hospital. In addition, the wiring configuration of the two devices obstructs patients’ daily activities. Consequently, a user-friendly and wearable device for heartbeat monitoring is of great potential in medical instrumentation innovation. Phtotoplethysmography (PPG) is a conventional technique for heart rate monitoring with reliable accuracy. As opposed to PPG, Bio-impedance analysis (BIA) offers a wider range of health-related parameter measurements. BIA is commonly used in body composition measurement, i.e. the estimation of fat, muscle, bone, and fluids levels in human body. Due to the noninvasive, low cost, and portability features of the bio-impedance analysis system, many researchers have conducted studies to apply it to a wider range of areas, such as physiological variables monitoring (e.g. heart rate [2] and respiratory rate), tissue state assessment [3], detection of ischemia [4] or cancer cells [5], etc. In cardiovascular studies, bio-impedance measurements enable the acquisition of information on stroke volume [6]. In [7] and [8], heart rate was detected from bio-impedance variations measured through the feet, by standing on a bathroom weighing scale intended for body composition measurement. These systems were easy to use but the configurations limited the possibility of continuous heart rate monitoring. Heart rate was measured on the wrist based on BIA with electrodes placed along the artery [2]. Such placement of electrodes was intended for pulse wave velocity measurement but not suitable for a wrist-worn application. In this paper, the authors aimed to design a user-friendly and wrist-worn device based on bio-impedance analysis for continuous heartbeat monitoring. The system applied electrodes on a subject’s wrist in a configuration that was perpendicular to the artery, which conformed to a wristband design. The system was tested against a reference on two subjects at two different locations on the wrist for heart rate detection. The results showed that the reported BIA wrist-worn monitoring system can effectively measure reliable heartbeat signals from the wrist with high accuracy. II. MEASUREMENT PRINCIPLES Biological tissues are primarily composed of cells and fluids. The equivalent electrical impedance model of blood cells is illustrated in Fig. 1. The resistance of the inner part of the blood cell is denoted in Ri, while Re represents the resistance of extracellular medium. The cell membrane behaves as a capacitor, which is represented in Cm. Due to the resistive and reactive behavior of each component, electrical

6294

is modulated by the impedance variation resulted from heartbeat pulses. With appropriate signal processing, the small variation of impedance ∆Z(t) can be obtained. III. BIO-IMPEDANCE MEASUREMENT SYSTEM DESIGN A. System Architecture The proposed architecture of bio-impedance measurement system is shown in Fig. 3. Instead of directly demodulating the output signal for ∆Z(t), a down-converting structure with a mixer is used, which is similar to the direct conversion in the radar receiver. A constant alternating current with an amplitude of 425µA and a frequency of 50kHz is injected through driving electrodes. The amplitude and frequency of the applying current is generated by a standard bio-impedance analyzer so that the system is able to detect low-level impedance variations without harming human subjects [10].

Figure 1. Electrical impedance model for blood including impedance variation related to heartbeat [2].

Figure 2. Wrist model and tetra-polar configuration for bio-impedance sensing, where red dots represent driving electrodes and black ones represent sensing electrodes [2].

The resistor R connected in series with human body is used to obtain the local oscillator (LO) signal. The reason for using a resistor of 100Ω is that the parallel connection between R and the input impedance of instrumentation amplifier (IA) has minimal effects on the actual voltage across the resistor R. Thus, the voltage of R can be represented as Vr(t) = RI0cos(ω0t)

current that passes through blood cells behaves differently according to frequencies. The current at low frequencies will flow mainly through the extracellular medium. At higher frequencies, the modeled cell membrane capacitor will act as shunt that allows current flow through the intracellular medium of blood cells. It has been reported that low-level heartbeat-associated impedance variations can be measured in limbs [9]. Heart rate thus can readily be detected from the wrist by measuring impedance variations due to the propagation of blood flow. As shown in Fig. 1, Z0 indicates the constant impedance of wrist tissues, and ∆Z represents the impedance variations directly related to heartrate. ∆Z can be extracted from a voltage signal across two sensing electrodes, illustrated as black dots in Fig. 2, on the wrist when a small constant alternating current is sent through two driving electrodes (represented as red dots). The voltage signal across the sensing electrodes is expressed as V (t) = (Z0 + ∆Z(t))I0cos(ω0t)

(1)

where I0cos(ω0t) is a small AC current with constant amplitude injected into the wrist. From (1) it can be seen that the measured voltage signal V(t) is a sinusoid whose amplitude

(2)

High-gain amplification is desired for both voltage signals because the impedance variation is on the order of mΩ and the corresponding voltage change is on the order of µV. Sufficient amplification of these two voltage signals can not only increase the accuracy of the measurement, but also effectively reduce the conversion loss of the mixer. The amplified Vr(t) and V(t) are sent to LO port and radio-frequency (RF) port of the mixer, respectively. The intermediate frequency (IF) signal is obtained at mixer IF port:

IF  G1RI 0cos(0t )  G2 ( Z 0  Z (t )) I 0cos(0t )  GR( Z 0  Z (t )) I 02cos 2 (0t ) 

G 2 G I 0 Z 0 R  I 02 Z (t ) R 2 2 G  I 02 R( Z 0  Z (t ))cos(20t ) 2

(3)

where G1 and G2 are the gains of the IA following the resistor R and the human body respectively, and G = G1G2. The bandpass filter is used to remove the DC offset related

Figure 3. Block diagram of heart-related impedance variation measurement.

6295

to basal impedance Z0 and the double-frequency component among mixer outputs. By sending the IF signal into a bandpass filter with corner frequencies of 0.3Hz and 3Hz, the heart-related signal S(t) can be obtained, which is G 2 G 2 I0 Z0 R  I 0 Z (t ) R 2 2 G 2  I 0 R ( Z 0  Z (t ))cos (20t ) 2 G 2  I 0 Z ( t ) R 2

S (t ) 

(4)

Since the impedance change related to heart pulse wave is very small, another high gain amplifier is necessary before S(t) is sent to the analog-to-digital converter (A/D). B. Experiment set-up Before turning the proposed bio-impedance measurement system into a full PCB circuit, an initial experiment was conducted to validate the feasibility and accuracy of this architecture for extracting heartbeat signals. The RJL Quantum II Bio-impedance Analyzer with four standard alligator clips that conform with the device setting was used as the AC source for desired constant alternating current. Four Covidien CA-610 electrodes were placed on the wrist for current driving and voltage detection. The electrodes were attached to the same side of the wrist, perpendicular to the artery, like a wristband configuration. Both upper and lower wrist were tested for validation. Fig. 4 shows the settings of electrodes. The pair at two ends induced AC current through the wrist, while the pair in the middle with 3.7cm separation from their centers were used for voltage sensing. A UFI Model 1010 piezoelectric finger pulse transducer [11] was attached on the index finger of the subject to obtain a reference pulse signal synchronously. To create LO source, a 100Ω resistor was connected in series with the current driving electrodes. The voltage signals from wrist and the resistor will have identical frequency and phase, satisfying the direct conversion principle. The amplification of the voltage signals were achieved by Stanford SR-560 low-noise amplifiers (LNA). They also performed filtering function in order to rule out noises in these two channels. For the human subject induced voltage signal

(RF), it was amplified by 50 times after a band-pass filter with cutoff frequencies at 10kHz and 100kHz. As for the resistor voltage (LO), it was amplified by 200 times and band-pass filtered at the same cutoff frequencies. Both of the signals were AC coupled in SR-560 to remove DC. After amplification, human subject’s voltage signal was fed to radio frequency (RF) port of a mixer (Mini-Circuits ZLW-6+), while that of the resistor was fed to LO port for mixing. The IF port outputs the mixed components to a band-pass filter (BPF). The function of the BPF was to remove double frequency components in the mixer products, preserving only the baseband signal containing heartbeat-related features. The cutoff frequencies of the BPF were 0.3Hz-3Hz, realized by another SR-560 LNA. The filtered signal was AC coupled as well, and amplified by 50 times to increase its resolution. The final heartbeat-related signal and reference were sent to NI USB-6218 for analog-to-digital conversion and acquisition. Further processing was performed in MATLAB. C. Human testing results The experiments were conducted according to the Committee on Human Studies (CHS) under protocol number 19176 and HRPO A-18177. In the conduct of research where humans are the subjects, investigators adhered to the policies regarding the protection of human subjects as prescribed by Code of Federal Regulations (CFR) Title 45, Volume 1, Part 46; Title 32, Chapter 1, Part 219; and Title 21, Chapter 1, Part 50 (Protection of Human Subjects). Two subjects were recruited in this study. Experiment results of subject #1 from lower wrist are presented in Fig. 5, where red traces are with the reference signal while blue traces represent BIA signals. It can be seen from Fig. 5(a) that there is power line noise in the raw bio-impedance signal. But it is evident that the envelope of the bio-impedance signal is in good agreement with the finger pulse reference. After performing filtering and peak detection in MATLAB, there are 35 peaks detected in both signals during the 30-second duration. Thus, the heart rate acquired from BIA system is 70BPM (beats per minute), matching its reference precisely. From the spectrum plots of the two signals shown in Fig. 5(c), it can be seen that accurate heart rate is also available, which is 1.16 ∗ 60 = 70BPM. The consistency of results from both time and frequency domain plots indicates that (1) heart rate can be monitored from the lower side of the wrist by measuring the heart-related impedance variation in a wristband configuration; (2) direct down-conversion structure of bio-impedance measurement system is feasible. The same experiment procedure was repeated on the upper side of wrist for the two subjects, which conformed with the same configuration as the lower wrist experiment. A summary of experiment results of subject #1 and #2 from both lower and upper wrist is listed in Table I, where heart rates detected from the proposed monitoring system are obtained and compared with finger pulse transducer results. It can be seen that the bio-

(a)

TABLE I.

(b)

Figure 4. Bio-impedance electrodes configuration on wrist at (a) lower side, and (b) upper side. Current driving electrodes are #1 and #4. Voltage sensing electrodes are #2 and #3.

6296

HEART RATE EXPERIMENT RESULTS Heart Rate Measured

Configuration

Subject

Bio-impedance Measurement System

Finger Pulse Transducer

Lower Wrist

#1

70BPM

70BPM

impedance variation at the wrist closely matched with the heartbeat signal acquired from a standard piezoelectric finger pulse transducer, which validated heart rate sensing by the proposed system. A general circuit design was also presented for a wearable heartbeat monitoring system. The PCB design and fabrication of the wearable will be completed in the future work. The consideration of acquiring more reliable data from larger population of human subjects and motion artifact cancellation will also be taken into account. The performance of other types of electrodes will be studied and evaluated in the future. (a)

ACKNOWLEDGMENT This research was supported by Archinoetics LLC. This work was supported in part by the US Army Medical Research and Materiel Command under Contract No. W81XWH-13-C-0027, and National Science Foundation (NSF) CBET-1160326. The authors would like to thank Rob Matthews from Archinoetics LLC for his general support. Special appreciation to Dr. Christopher Stickley from Kinesiology and Rehabilitation Science department, University of Hawaii at Manoa for loaning the BIA analyzer device and guidance.

(b)

REFERENCES [1] American Heart Association, ”Heart Disease and Stroke Statistics,” 2015.

[2] M.-C. Cho, J.-Y. Kim, and S. Cho. “A bio-impedance measurement

[3]

[4] (c) Figure 5. Experiment results of subject #1 from lower wrist, where red traces represent finger pulse signal and blue traces represent bio-impedance signal. (a) Time-domain plots of recorded raw signals. (b) Time-domain plots of filtered signals. (c) Frequency-domain plots of filtered signals

Configuration

Upper Wrsit

[5]

[6]

Heart Rate Measured

Subject

Bio-impedance Measurement System

Finger Pulse Transducer

#2

64BPM

64BPM

#1

70BPM

70BPM

#2

64BPM

66BPM

[7]

[8]

impedance measurement system yields reliable results that agree with the reference.

[9]

IV. CONCLUSION

[10]

This work presented a bio-impedance analysis based heart rate monitoring system. The design used only four electrodes to sense heartbeat related signals on the wrist where wristband is normally placed. Measurements were taken with the proposed setting and analyzed in time domain as well as on frequency spectrum. The measured results showed that

[11]

6297

system for portable monitoring of heart rate and pulse wave velocity using small body area,” IEEE International Symposium on Circuits and Systems (ISCAS), pp. 3106-3109, 24-27 May, 2009. L. Nescolarde, J. Yanguas, et al, ”Assessment and follow-up of muscle injuries in athletes by bioimpedance: Preliminary results,” IEEE 33rd Annual International Conference of the EMBS, Boston, Massachusetts, USA, Aug. 30 - Sept. 3, 2011 Fritz Mellert, Kai Winkler, et al, ”Detection of (Reversible) Myocardial Ischemic Injury by Means of Electrical Bioimpedance,” IEEE Transactions on Biomedical Engineering, vol.58, no. 6, June 2011 A. R. Al-Hashimi, A. N. Nordin, and A. W. Azman, ”Bioimpedance spectroscopy system for Characterization of Cancer Cells,” 5th International Conference on Intelligent and Advanced Systems (ICIAS), Kuala Lumpur, June 2014 S. M. M. Naidu, Uttam R. Bagal, Prem C. Pandey, et al, ”Monitoring of Stroke Volume through Impedance Cardiography Using an Artificial Neural Network”, 21st National Conference on Communications (NCC), Mumbai, Feb. 27 - March 1, 2015 Padma Batra1 and Rajiv Kapoor, ”A Novel Method For Heart Rate Measurement Using Bioimpedance,” International Conference on Advances in Recent Technologies in Communication and Computing, Kottayam, 16-17 Oct. 2010 Delia H. Diaz, Ó Casas, and Ramon Pallas-Areny, ”Heart Rate Detection from Single-Foot Plantar Bioimpedance Measurements in a Weighing Scale”, IEEE 32nd Annual International Conference of the EMBS, Buenos Aires, Argentina, August 31 - September 4, 2010 T. M. R. Shankar, J. G. Webster, and S.-Y. Shao, “The Contribution of Vessel Volume Change and Blood Resistivity Change to the Electrical Impedance Pulse”, IEEE Transaction on Biomedical Engineering, vol. 32, no. 3, March 1985 J. G. Webster, Medical Instrumentation: Application and Design, Wiley & Sons, 3rd ed., 1997 Model 1010 Piezoelectric Pulse Transducer Manual, UFI, Morro Bay, CA, 2010.