Journal of Medical and Biological Engineering, 20(4): 193-202 2000

193

Portable Impedance Cardiography System for Real-Time Noninvasive Cardiac Output Measurement Liang-Yu Shyu*

Chia-Yin Chiang

Chun-Peng Liu1

Wei-Chih Hu

Department of Biomedical Engineering, Chung-Yuan Christian University, Chung-Li, Taiwan, 320 ROC 1 Department of Cardiology, Veterans General Hospital- Kaohsiung, Kaohsiung, Taiwan, 800 ROC Received 3 August 2000; Accepted 27 September 2000

Abstract Impedance cardiography (ICG) is the only known method that can observe the continuous left ventricular volume changes non-invasively and conveniently. This paper proposes to develop a portable ICG system for long-term continuous cardiac activity monitoring using the impedance technique. This ICG system combines the advance analog amplifier with the calculation power of digital signal processing chip to achieve real-time monitoring. One of the major differences between the proposed system and the existing systems is in the bandwidth of dZ/dt circuit. Using proper bandwidth can significantly increase the reliability and validity of ICG signal. Additionally, the proposed system is light weight and suitable for a portable monitoring system.The results demonstrate that the proposed ICG system is capable of tracking the dynamic changes of SV and CO during physiological challenges such as Valsava maneuver and the hand grip task. On the other hand, the results also show that there is no difference between CO obtained by the proposed ICG system and the Fick method. Additionally, the correlation coefficient is 0.91. The results indicate that the proposed system is a valid and reliable tool for patient monitoring and it is capable of assessing the dynamic cardiac activity information.Due to the simplicities of the current signal processing and events detection methods, the CPU is idled for substantial amount of time. Thus, in the future, more advanced techniques such as adaptive filtering and wavelet analysis can be implemented for event detection to increase the accuracy and reliability. Keywords：Impedance cardiography, Cardiac output, DSP

Introduction Cardiovascular disease is one of the leading causes of mortality in most of the developed countries. Although invasive techniques, such as thermodilution, give satisfactory results on the estimation of stroke volume (SV), they are not suitable for long-term or repeated measurements because of their invasive nature. During the past two decades, several non-invasive techniques capable of monitoring cardiac activity are available, including ultrasound Doppler, magnetic resonance imaging, and impedance cardiography (ICG). However, all these techniques except impedance cardiography are not suitable for long-term continuous monitoring of cardiac activity. The impedance cardiography has been under intensive investigation since 1960's. The relationship between resistance and the volume change of a homogeneous cylindrical conducting material is frequently used as the basis of cardiac impedance plethysmography. Through a simple mathematical exercise [1], it can be shown that an increase in the volume of the conducting material is accompanied by a decrease in the * Corresponding author: Liang-Yu Shyu Tel: +886-3-4563171 ext.4500; Fax: +886-3-4563171 ext.4599 E-mail : [email protected]

resistance, when uniform electrodes. Additionally, when this technique current density distribution exists between the measuring is applied to SV estimation, Kubicek [2] developed the forward extrapolation method to compensate for the blood flowing out of the measurement region and obtained the well-known Kubicek equation for SV estimation. In addition to the SV, parameters such as cardiac output (CO), myocardial contractility, and systolic time intervals (STIs) can also be derived from the measured impedance signal. Numerous studies [3-5] have been conducted to examine the reproducibility, accuracy and validity of the ICG. Although a few conflicting results were reported, the ability of ICG to track relative changes of SV and CO is generally appreciated. On the other hand, the absolute value of SV obtained by ICG is highly controversial. Previous studies suggest that various physiological variables contribute to the impedance waveform. The four most commonly mentioned contributors are: 1) blood volume change in the heart [6,7]; 2) change of blood volume in the aorta and vena cava [6-8]; 3) changes of blood conductivity due to the orientation of red blood cells in flowing blood [7,9]; and 4) lung resistivity change [6,10,11]. It is generally agreed that the ICG originates from a combination of blood volume change, rearrangement of current conductors, redistribution of

194

J. Med. Biol. Eng., Vol. 20. No. 4 2000

current density, and resistivity change of flowing blood within the measuring segment. There is no agreement, however, on the amount of contribution from each factor to the impedance signal at a particular time because the current distribution within the thorax is virtually impossible to model and study [1]. Numerous validation studies comparing the ICG with other SV estimation techniques have been conducted in the past years. Most of them result in high correlation and demonstrate that ICG is a reliable method in monitoring cardiac activity. To further explore the potential of ICG, a portable system for long-term continuous CO monitoring is proposed. In addition, the performance of this new patient monitoring system is evaluated using Fick method. In the field of impedance cardiography, the tetrapolar electrode system is the most widely used electrode system. A low intensive high frequency constant current is injected into the subject’s thorax through the two outer electrodes. The two inner electrodes sense the small impedance change, during each cardiac cycle. The impedance signal is then amplified, processed and used in the estimation of SV. Several different placements of electrodes have been proposed in the past few years. However, the electrode arrangement proposed by Kubicek et al. [2] is still the most widely accepted. On the other hand, in the past few years, some investigators used spot electrodes to replace the traditional band electrode [5]. There are two advantages to the spot-electrode system, i.e., it is easier to apply and is more comfortable than the band electrodes. Additionally, there is no visible difference in the impedance signal from that obtained by the band-electrode system [12-14]. Thus, in this project, spot electrode system w a s u s e d . #1

#4

Current Protection circuit

#2

#3

Pre - Amplifier (AMP 01)

Constant Current Source 100KHz 4mA(RMS)

80KHz High Pass Filter

Over Load Alarm Circuit

120KHz Low Pass Filter

Power Supply System

Amplitude Demodulation Circuit

9.6V Ni-Cd Battery

Differentiator

4.26 Hz High Pass Filter

25 Hz Low Pass Filter

Constant Current Source 100kHz 4mA(RMS)

Impedance Cardiography

#2

#1

dz/dt Zo

#3

#4 ECG Circuit (R-wave Int.)

Personal Computer

l

RS232 Interface TMS320C50 DSP Chip

A/D-D/A Converter

8255

4*3 Key Board

128 * 64 dot LCD Display

32K SRAM

32K ROM

Figure 2. The block diagram of ICG system. The system includes： 100KHz constant current source﹑the analog circuits for ICG, ECG R-Wave detection circuit﹑A/D converter, digital signal processor (DSK), 8255 interface circuit, RS232 interface circuit, external memory, a 4-by-3 keyboard and an LCD display. A 100KHz﹑4mA(rms) constant current is generated by the constant current source and injected into the subject’s thorax through two outer electrodes. The thoracic impedance changes are amplified, demodulated and filtered to obtaine the dZ/dt and Z0. The DSP module determines the characteristic points in the ICG signal and obtains the SV estimation using Kubicek equation. The results are displayed on the LCD display. In addition, they can be transferred to personal computer through the RS232 interface circuit.

Methods

Top of R-Wave Trigger Circuit

ECG Int.

0.2 Hz Low Pass Filter

Zo

0.1 Hz High Pass Filter

40 Hz Low Pass Filter

∆Z

Amplifier

30 Hz Low Pass Filter

dz/dt

Full-wave Rectifier

Figure 1. The block diagram of ICG circuit. To obtain the dZ/dt, Z0, and ∆Z, the thoracic impedance change, obtained from electrode 2 and 3, is amplified, demodulated, and filtered.

There are three major blocks in an ICG system: the high-frequency constant current source, the dZ/dt amplification circuit, and calibration circuit. Figure 1 depicts the block diagram of the proposed circuit. One of the major differences between the proposed system and the existing systems is in the bandwidth of dZ/dt circuit. It was shown that the appropriate upper bandwidth of the dZ/dt signal is in the range of 45 to 55 Hz [15]. Using improper bandwidth can cause significant signal distortion that may influence the validity of ICG. In addition to the analog circuit for extracting impedance signal, the proposed system also include: the analog to digital converter, the digital signal processing module (DSK, Texas Instrumentation Co.), the peripheral devices, the system power supply, the data storage memory, and the RS232 interface circuit (Figure 2). In the field of impedance cardiography, the Kubicek equation (Equation1) is most commonly used for SV estimation

Portable ICG for CO Measurement

SV =

ρL2 Z 02

× LVET × dZ dt

(max)

(1)

where ρ is the resistivity of blood in Ω-cm, L is the electrode distance between the two sensing electrodes in cm, Zo is the basal impedance in Ω, dZ/dtmax is the maximum rate of change of the impedance in Ω/s, and LVET is the left-ventricular-ejection time in s. System description In general, ICG system uses high frequency (10KHz ~ 100KHz) and low intensive (below 10mA rms) constant current source. Considering the resolution and patient safety, a 100KHz, 4mA (rms) constant current source is used in this system. The constant current source of the proposed system includes an overflow alarm and transient protection circuits. When there is a bad connection on the two outer electrodes, the output voltage is higher than a pre-determined value, a warning signal is triggered to light an LED and warn the user. Furthermore, when the device is turned on, the constant current source is unstable. Thus, to protect the subject, this system employs a delay device in front of the output terminal to ensure that the current is not delivered until one second after the system is turned on. The block diagram of the proposed ICG system is illustrated in Figure 1. An instrumentation amplifier amplifies the modulated impedance signal that is extracted by the two inner electrodes. It then passes through an 80 kHz highpass filter followed by a 120 kHz lowpass filter before it is demodulated. On the other hand, it has been shown that the bandwidth of the impedance signal is in the range from DC to 49 Hz for normal subjects with a heart rate (HR) below 120 beats-per-minute (bpm) [15]. Thus, a 100Hz lowpass filter is used to reduce the high frequency components that are generated by the half-wave-rectifier demodulation circuit. After demodulation and amplification, the thoracic impedance signal Z can be separated into two distinct parts. The time varying component designated as ∆Z is obtained by highpass filtering the thoracic impedance signal. On the other hand, the near DC component is known as Z0. To estimate the SV using the Kubicek equation, however, it is the maximum rate of change of impedance is required. Therefore, an additional dZ/dt signal extracting circuit is also needed. To generate dZ/dt signal, the thoracic impedance signal is differentiated directly in order to achieve better low frequency characteristics. The differential circuit works like a highpass filter with very small DC gain. The DC component of thoracic impedance signal is heavily attenuated. Since this system is designed for using by subjects with HR below 120 bpm, the bandwidth of dZ/dt signal is limited to below 30Hz [15] using a fourth-order Butterworth lowpass filter. Since the cardiac systolic events occur after the onset of QRS complex in the ECG signal, it is convenient to use the R wave as a synchronous signal for SV calculation. The ECG signal obtained from the two inner electrodes is used as the input to this circuit. The waveform of the extracted ECG signal is far from the standard. However, it is the temporal position of the R-wave that is important in the generation of the interrupt

195

signal. Therefore, a standard ECG signal is not required. In the proposed system, a negative pulse is generated upon the occurrence of an R wave occurs, as an interrupt signal to the DSP chip to start the SV estimation. The proposed system uses the TMS320C50 (C50) DSP Starter Kit (DSK, Texas Instruments Co.) as the heart of digital signal processing module. The C50 is a 16 bits fixed point digital signal processor. It can perform 20 million instructions per second (MIPS). In addition to the DSP chip, the DSK contains 10K words of RAM, two serial ports, an A/D and D/A converter. Its maximum addressable external memory space is 224K words including 64K words for program, 64K words for data, 64K words for I/O, and 32K words for global storage. In addition to the DSK, several external circuits are included for display and communication. To overcome these restrictions of not having enough memory space for storage and processing, 32K words of SRAM and 32K bytes of EPROM are included in the proposed system. In addition, an 8255, a 74922, a 128-by-64 dot matrix LCD, a four-by-three keyboard and an RS232 interface circuit are combined as the external circuit. The C50 is set in the microprocessor mode in order to operate as a stand-alone device. When the system is reset, the CPU loads in instructions that are stored in the EPROM by using the C50 boot loader. To display the results and interface between system and the user, a 128-by-64 dots matrix LCD is used. Informations such as CO, SV, HR, LVET, as well as diagrams of lone-term CO and HR trends are displayed on the LCD. In addition, system statuses are displayed while the user communicates with the system using a keypad. Since the speed of LCD is much slower than C50, an 8255 chip is needed as the interface between it and C50. A four-by-three keypad is used to setup the system and input parameters. The interface of this keypad is achieved by using a 74922 chip. The 74922 chip sends an interrupt signal to DSP when a key is pressed. On the other hand, the DSP chip gets the information of the pressed key through the 8255. The port A of this 8255 chip is used to connect with the data port of LCD. On the other hand, ports B and C are used to connect with the keyboard interface IC and the control line of LCD, respectively. The proposed system is equipped with an RS232 communication function. Through the RS232 interface circuit, the information can be transmitted and saved in the PC using a commercial software after each experiment for further analysis. Furthermore, to achieve a portable monitoring system, a single 9.6V 650mA Ni-Cd battery is used to power the proposed system. Calculation and signal processing The software is written in TMS320 fixed-point DSP assembly language. When the system is reset, program is downloaded from the EPROM. After the booting procedure, system initializes the memory, system, and analog interface circuit (AIC). Memory of the proposed system is configured as follows: 1K words is set as data section, 4K words as data section, 5K words as program section and the 32K words of external memory is set as external data section.

196

J. Med. Biol. Eng., Vol. 20. No. 4 2000

Two signals are digitized, the dZ/dt and Zo in a sampling rate of 1000Hz (500Hz per channel). Additionally, the interleave method is used that is when one sample from channel one (dZ/dt) is being converted and the interrupt handler routine changes the input channel from channel one to channel two. Thus, the signal from channel two will be converted when the next interruption occurs. The digitized dZ/dt and Zo data are saved in two different 1K-words circular queues. For a sampling rate of 500Hz, two seconds of data can be stored in these circular queues. In general, this is enough for real time ICG processing. On the other hand, since Zo is a slow changing signal, the proposed system uses only four samples of Zo for calculation. The user can communicate with the system through a four-by-three keypad. When the keyboard is pressed, the keyboard interfacing IC relays the code and issues an interrupt to C50. The interrupt service routine reads in the code from the port B of 8255 and saves to a register. The keyboard interrupt service routine sets a flag to inform the main routine to fetch the keyboard information. The ECG R-wave interrupt signal plays a very important role in the signal processing program. Since the CO could only be calculated after the cardiac cycle is completed. Therefore, at each R-wave interrupt, the interrupt service routine set an interrupt flag and marks the data address where interruption occurred so that the main program can finish the calculation of CO.

ECG

PCG

component within the second heart sound, respectively, when it is compared with the phonocardiogram [16]. After the ECG interrupt is received, the main signal processing routine first detects events such as the peak of dZ/dt during systolic, B point, and X point. Parameters such as HR, LVET and dZ/dtmax are then calculated for subsequent use in the estimation of SV. The time interval between two ECG R-wave interrupts is converted into HR. On the other hand, it is known that all the characteristic points (B, Z, and X points) locate within the electromechanical systolic (QS2) interval. To speed up the detection routine and limit the search interval, the QS2 interval is first estimated by the van Der Hoeven equation: QS2 (in ms)= 501 - 1.61 * HR. The Z point is designated as the maximum of dZ/dt signal during the systolic period. It can be found within a 256 ms interval after the ECG R-wave interrupt. Points B and X are identified as the onset of the rapid upstroke and the lowest point in the dZ/dt waveform, respectively [17]. The B point in the dZ/dt signal represents the time the large outflow of blood from the left ventricular and is defined as the time when rapid up stroke in the dZ/dt signal occurs. Accordingly, after second differentiation, B point can be located as a local maximum. A second order digital differentiator using five points difference equation is used in the current system:

Y

,,

ٛ= Y 0 − 2 Y − 2 + Y − 4

On the other hand, it is very difficult to detect X point from the ICG signal because it has no distinct characteristic. The X point is defined as the local minimum at the end of systolic. Therefore, an interval of 32ms before and after the end of estimated QS2 is searched for local minimum. After the peak of dZ/dt, B, and X points are located, the dZ/dtmax is determined as the amplitude difference between the B point and the Z point, as proposed by Mohapatra [18]. Additionally, the LVET is defined as the time interval between B and X points. The SV is then calculated according to the Kubicek equation and the CO is calculated as:

CO = SV × H R dZ/dt

Figure 3.

Events related to the impedance cardiography.

Event detection Three events in the ICG, i.e., the B point, the maximum of dZ/dt during systole, and the X point, are used in the determination of dZ/dtmax and LVET (Figure 3). The B and X points are the local minima before and after the principal peak in the dZ/dt tracing. The LVET is the time interval between the aortic valve opening and closure. In the dZ/dt waveform, B and X points are found to correspond with the maximal vibration of the first heart sound observed at the apex and aortic-closure

(2)

(3)

After calculation, parameters such as SV, HR, CO, and LVET as well as the long-term trends are plotted on the dot matrix LCD. Limited by the external memory space, the current system does not record all the raw data but only the results of SV, HR, Zo, dZ/dtmax and time intervals of detected events. Five words of data are recorded per every cardiac cycle. For a subject with heart rate of 80 bpm, the proposed system can store 80 minutes of ICG data. After the experiment, the stored data is transmitted to a personal computer via the RS232 communications port. Statistical analysis The Pearson correlation coefficient, Student t and paired t tests will be used for between and within groups comparison for all parameters. All results with a value of p less 0.05 will be regarded as statistically significant.

Portable ICG for CO Measurement

197

0 .1 2 0 .1 0 .0 8 0 .0 6 0 .0 4 0 .0 2 0 -0 .0 2 -0 .0 4 -0 .0 6

Subject

Four patients from the Kaohsiung Veterans General Hospital were recruited to evaluate the system performance. Subjects with severe valvular dysfunction, marked renal dysfunction, big left ventricular aneurysm, marked emphysema, or prominent congestive heart failure were excluded. Inform (a ) N o rm a l B re a th in g consent was signed prior the study. Additionally, all subjects 0 .1 2 Portable ICG for Real-time CO Measurement 0 .1 had normal sinus rhythm. 0 .0 8 Before the routine diagnostic catheterization, four 0 .0 6 Ag/AgCl electrodes were placed on patient’s upper thorax. The 0 .0 4 0 .0 2 placement of electrodes was a modified version of that 0 proposed by Wang et al. [19]. Lead 1 was placed on the back -0 .0 2 of upper neck. Lead 2 and lead 3 were placed on the base of -0 .0 4 -0 .0 6 the neck in the midaxillary line on the left side of the neck and (b ) S to p B r e a th in g 0 .1 2 the xiphisternal notch, respectively. Lead 4 was placed on the 0 .1 abdomen at least 3 cm below lead 3. The outer two electrodes 0 .0 8 were used as the current injecting electrodes. The inner two 0 .0 6 0 .0 4 electrodes were used as the signal sensing electrodes. 0 .0 2 After the evaluation of the patient’s clinical problems, 0 -0 .0 2 subjects were asked to rest for five minutes followed by -0 .0 4 handgrip task and Valsava maneuver with two minutes rest -0 .0 6 (c ) B re a th R u s u m e d period in between the two tasks. During the trial, three CO’s were estimated: before the end of resting period, one minute into the handgrip task, and at the end of Valsava maneuver, Figure 5. Typical recording of dZ/dt signals during (a) normal using the Fick principle. breathing, (b) breath holding and (c) breath resumed for a normal subject.

I(R M S )

Through out the trial period, the ICG event detection and SV estimation are performed in real time by the proposed system. In addition, dZ/dt signal with event markers is acquired by a personal computer for further evaluation of the accuracy of the event detection routines.

4 .2 0

mA(rms)

4 .0 0 3 .8 0 3 .6 0

Results

3 .4 0

In this section, the results of the hardware performance are presented followed by the validating results. Figure 4 depicts the test result of the constant current source. It is clearly demonstrated that the output of this current source is constant even when the load is more than 500 Ω. The typical recording of dZ/dt signals during normal breathing, breath holding and breath resumption, from a normal subject, are illustrated in Figure 5. Since the dZ/dt signals from this proposed ICG system are amplified and filtered with proper bandwidth, the fidelity of ICG signals are clearly preserved. Not only the signal-to-noise ratio is high but also the characteristic points such as B and X points are clearly visible which is much helpful for event detection. On the other hand, the outputs of the system including parameters such as SV, HR, CO, and the long-term trends are plotted, in real-time, on the dot matrix LCD (Figure 6).

3 .2 0 3 .0 0 0

500

1000 ohm

1500

2000

Figure 4. The test result of the 100KHz 4mA(rms) constant current cource. The Y axis is the output current in mA and the X axis is the load resistance in ohm.

198

J. Med. Biol. Eng., Vol. 20. No. 4 2000

120 100 HR (b p m )

(a)

80 60 40 20

120 100

(b)

80

SV (m l)

60 40 20

8

Figure 6. The LCD display. In addition to the SV, CO, HR and electrode distance, the long-term trends of (a) CO and (b) HR are also displayed.

6 CO 4 (L /m in ) 2

Start

End

0 0

30

60

90

120

150

1 8 0 (S )

Figure 7. The responses of HR, SV, and CO during the Valsava maneuver from a health young male subject. The data are obtained using the proposed system. Table 1. The results of ICG system and Fick principle. The results of ICG system are averaged over 10 heartbeats

Patients

A

Blood Fick CO (L/min) ICG CO (L/min)

B

C

1

2

3

1

2

3

1

2

3

1

2

3

5.0

5.5

5.0

5.4

5.1

5.2

3.7

3.1

3.5

4.2

4.1

3.5

5.51 ± 4.88 ± 5.19 ± 4.94 ± 5.44 ± 5.38 ± 1.44 ± 3.18 ± 3.45 ± 0.43 ± 0.48 ± 2.14 ± 0.715 0.728 0.214 0.862 0.363 0.453 0.139 0.577 0.896 0.005 0.005 0.41

120

120

100

100

H R (b p m )

D

H R (b p m )

80

80 60

60 40

40

20

120

20 100

100

80

S V (m l)

S V (m l)

60

80 60 40

40

20

20

8

0 8

6

6

C O ( L / m i n )4 2

C O 4 ( L /m in )

Start

End

0

2

St ar t 3 0s

0

En d 55 s

0 0

30

60

90

120

150

1 8 0 (S )

Figure 8. The responses of HR, SV, and CO during the Valsava maneuver from a patient. The data are obtained using the proposed system

30

60

90

120

150

1 8 0 (S )

Figure 9. The responses of HR, SV, and CO during the handgrip task from a normal young male subject. The data are obtained using the proposed system

The Valsava Maneuver The Valsava is known to cause a rapid decrease in SV at the beginning of the maneuver. When the breath is resumed, the SV increase immediately with a substantial overshot before it returns to normal resting level. On the other hand, the HR increases at the beginning of this maneuver that compensates the decrease in SV. Consequently, the CO’s are maintained at a

Portable ICG for CO Measurement

during the rest period in patient C. On the other hand, poor signal quality from patient D caused the CO’s estimated from ICG system significantly lower than normal. After excluding these four data points, there is no difference between these two data sets (Figure 11). Additionally, the correlation coefficient is as high as 0.91.

120 100

H R (b p m )

199

80 60 40

20 120

Discussion

100

SV (m l)

80

According to the Kubicek’s equation, the SV is directly proportional to LVET and dZ/dtmax. The detection of B, Z, and X points in turn affects the correctness of LVET and dZ/dtmax. Figure 12 illustrates dZ/dt signals obtained from the four patients. It can be seen that the signal amplitude from patient D

60 40 20 10

8 C O 6 (L /m in )

7 4 Start 30s

6

End 90s

2 0

30

60

90

120

150

1 8 0 (S )

Figure 10. The responses of HR, SV, and CO during the handgrip task from a patient. The data are obtained using the proposed system.

5 CO 4 (L/min) 3

Fick ICG

2 1

fairly constant level with two visible changes at the beginning and at the end of the maneuver. A typical result of the Valsava maneuver from a healthy young male subject, using the proposed system, is illustrated in Figure 7. Only one out of four patients shows similar but less dramatic changes as compared to normal subjects (Figure 8). At the beginning of the maneuver the SV decreased, however, there is no overshot at the end of Valsava. Additionally, for all patients, there is no significant increase in HR during Valsava maneuver.

0 A1

A2

A3

B1 B2 B3 Patient (test point)

C2

C3

Figure 11. Results after exclude four data points, one data point from the patient C and three data points from patient D. There is no different between these two data sets. The correlation coefficient is as high as 0.91. The results of ICG system are averaged over 10 heartbeats.

P a tie n t (A )

The Hand Grip Task The handgrip is known to increase HR and decrease CO during the task. Figure 9 illustrates a typical result of handgrip task from a healthy young subject obtained by the proposed system. The SV decreases starting from the beginning of the task. The change is less dramatical when compared with the response of Valsava maneuver. The CO decreases slightly during the task. Only one of the four patients showed similar responses on SV and CO. However, the heart rate did not increase during the handgrip task (Figure 10). It is clearly demonstrated that the proposed ICG system can reveal the dynamic change.

P a tie n t (B )

P a tie n t (C )

P a tie n t (D )

Figure 12. dZ/dt signals obtained from four different patients.

Fick principle Under the assistance of physicians, there were three CO measurements using Fick principle for each patient. The experimenter entered the time of blood drawing. At the end of each experiment, the 10 beat-to-beat CO, from the ICG system, before each blood drawing were averaged and used to compare with the Fick method. The results are illustrated in Table 1. There is significant difference between these two sets of data (p < 0.1). In addition, the Pearson correlation coefficient is low (r = 0.685). After carefully reviewing the recorded signals and the logbook, it is found that there was a bad-connection in one of the electrodes

is noticeably lower than the signals from other patients. This abnormal signal causes Z point detection errors and results in significantly lower SV estimation. The current signal processing and events detection methods are based on the results of the previous studies in this laboratory. They are simple and easy to implement. However, it is found that the noise in the incoming signal and the abnormal signal causes detection errors especially in the X point determination. Although, in this study, different lengths of searching period were used to reduce the detection error, the X point detection errors are still common. However, the error that it generated is not more than 20ms which is 5 % in the

200

J. Med. Biol. Eng., Vol. 20. No. 4 2000

LVET estimation. Previous studies demonstrate that, when comparing the ICG with thermodilution, the reported correlation coefficients between these two methods range between 0.83 and 0.88 for normal subjects [20,21] and heart patients [20,22-25]. Some studies reported that there is no significant difference between these two methods [20,23], whereas others claimed that impedance cardiography measurement reveals greater dispersion [25], or ICG underestimates SV [24], or ICG overestimates SV at very low flow rates [1]. Additionally, Demeter et al. [26] reported that in dogs no significant difference between impedance and thermodilution at the baseline is observed but some significant differences are found during inspiration and expiration. Both direct and indirect Fick methods were compared with the ICG. Edmunds et al. [27] found in nine adults and 14 children at rest and during steady exercise, that there is a close correlation between the ICG and the indirect Fick method with a correlation coefficient 0.94. High correlation coefficients were also reported by Miyamoto et al. [28], Teo et al. [29], and Zhang et al. [30] with correlation coefficients 0.91, 0.93, and 0.96 for healthy males, patients, and normal subjects, respectively. Enghoff and Lovhein [31] reported that, for patients, the absolute values of SV estimated by the impedance and the direct Fick method have no reliable agreement. They also concluded that CO cannot be accurately determined by the ICG for patients with advanced heart disease. However, the result of this study shows that the proposed system obtain the absolute values of SV from patients and the values are highly correlated with those obtained by Fick’s method. The proposed system also demonstrates the ability of continuously tracking the SV during different maneuvers.

Conclusion In this paper, a portable impedance cardiography system for real-time noninvasive cardiac output monitoring is developed and evaluated. This ICG system combines the advance analog amplifier with the calculation power of digital signal processing chip to achieve real-time monitoring. It is capable of detecting events in the incoming signal under different circumstances. Additionally, its ability to track the dynamic changes of SV and CO during physiological challenges is clearly demonstrated. The current signal processing and events detection methods are based on the results of previous studies by the author and his colleagues [32-34]. Due to their simplicities, the CPU is idled for substantial amount of time. Thus, in the future, more advanced techniques such as adaptive filtering and wavelet analysis can be implemented for event detection to increase the accuracy and reliability. In the area of hardware improvement, two suggestions are made. The first is to use a larger storage device such as PCMCIA flash memory card for long period of data and result storage. Second, by using fewer components and more power saving devices, it is expected that a 24-hour monitoring system can be achieved in the future. In conclusion, the proposed system can track the dynamic changes of SV, HR, and CO. Thus, physicians can study the

cardiac performance of patients with myocardium infarctions in a normal setting. On the other hand, the patient can benefit from this long term monitoring device that will provide real-time warning for lift-threatening situation in the future.

Reference [1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9] [10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

L. A. Geddes, and L. E. Baker, Principle of Applied Biomedical Instrumentation Third Edition, John Wiley & sons, New York, 1989. W. G. Kubicek, J. N. Karnegis, R. P. Patterson, D. A. Wrrsoe, and R. H. Mattson, “Development and Evaluation of an Impedance Cardiac Output System,” Aerospace Medicine, 1208-1212, December 1966. D. P. Bernstein, “Continuous Noninvasive Real-time Monitoring of Stroke Volume and Cardiac Output by Thoracic Electrical Bioimpedance,” Critical Care Medicine, 14(10): 898-901, 1986. W. G. Kubicek, F. J. Kottke, M. U. Ramos, R. P. Patterson, D. A. Witsor, J. W. Labree, W. Remole, T. E. Layman, H. Schoening, and J. T. Garamella, “The Minnesota Impedance Cardiograph - Theory and Applications,” Biomedical Engineering. 9(9): 410-416, September, 1974. M. Muzi, T. J. Ebert, F. E. Tristani, D. C. Jeutter, J. A. Barney, and J. J. Smith, “Determination of Cardiac Output Using Ensemble-Averaged Impedance Cardiograms,” J. Appl. Physiol., 58(1): 200-205, 1985. K. Sakamoto, K. Muto, H. Kanai, and M. Iizuka, “Problems of Impedance Cardiography,” Med. & Biol. Eng. & comput., 17, 697-709, 1979. D. W. Kim, L. E. Baker, J. A. Pearce, and W. K. Kim, “Origins of the Impedance Change in Impedance Cardiography by a Three-Dimensional Finite Element Model,” IEEE Transactions on Biomedical Engineering, 35(12): 993-1000, December 1988. W. G. Kubicek, “On the Source of Peak First Time Derivative (dZ/dt) During Impedance Cardiography,” Annals of Biomedical Engineering, 17: 459-462, 1989. R. Lambers, K. R. Visser, and W. G. Zijlstra, Impedance Cardiography, Van Gorcum, Assen, The Netherlands, 1984. R. P. Patterson, “Sources of the Thoracic Cardiogenic Electrical Impedance Signal as Determined by a Model,” Med. & Biol. Eng. & Comput., 23: 411-417, 1985. J. Kosicki, L. Chen, R. Hobbie, R. Patterson, and E. Ackerman, “Contributions to the Impedance Cardiogram Waveform,” Annals of Biomedical Engineering, 14: 67-80, 1986. B. B. Sramek, “Noninvasive Technique for Measurement of Cardiac Output by Means of Electrical Impedance,” Proceeding of the Vth ICEBI, Tokey, 39-42, Aug, 1981. M. Qu, Y. Zhang, J. G. Webster, and W. J. Tompkins, “Motion Artifact From Spot and Band Electrodes During Impedance Cardiography,” IEEE Transactions on Biomedical Engineering, 33(11): 1029-1036, 1986. B. C. Penney, N. A. Patwardhan, and H. B. Wheeler, “Simplified Electrode Array for Impedance Cardiography,” Med. & Biol. Eng. & Comput., 23: 1-7, 1985. B. E. Hurwitz, L. Y. Shyu, C. C. Lu, S. P. Reddy, N. Schneiderman, and J. H. Nagel, “Signal Fidelity Requirements For Deriving Impedance Cardiographic Measures of Cardiac Function Over a Broad Heart Rate Range,” Biological Psychology, 3-21, 1993. Z. Lababidi, D. A. Ehmke, R. E. Durnin, P. E. Leaverton, and R. M. Lauer, “The First Derivative Thoracic Impedance Cardiogram,” Circulation, XLI: 651-658, April, 1970. A. Sherwood, M. T. Allen, J. Fahrenberg, R. M. Kelsey, W. R. Lovallo, and L. J. O. van Doornen, “Methodological Guidelines for Impedance Cardiography,” Psychophysiology, 27(1): 1-23, 1990.

Portable ICG for CO Measurement

[18] [19]

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27]

S. N. Mohapatra, Non-invasive Cardiovascular Monitoring by Electrical Impedance Technique. Pitman Medical, 1981. X. Wang, X., H. H. Sun, D. Adamson, and J. M. van De Water, “An Impedance Cardiography System: A New Design,” Annals of Biomedical Engineering, 17: 535-556, 1986. R. W. Gotshall, V. C. Wood, and D. S. Miles, “Comparison of Two Impedance Cardiographic Techniques for Measuring Cardiac Output,” Annals of Biomedical Engineering, 17: 495-505, 1989. M. Muzi, D. C. Jeutter, and J. J. Smith, “Computer-Automated Impedance-Derived Cardiac Indexes,” IEEE Transaction on Biomedical Engineering, 33(1): January, 1986. K. E. Bloch and E. W. Russi, “Comparison of Impedance Cardiography to Invasive Techniques for Measurement of Cardiac Output,” AM J Cardiol, 79(6): 846, 1997. W. V. Judy, D. J. Powner, K. Parr, R. Demeter, C. Bates, and S. Marshall, “Comparison of Electrical Impedance and Thermal Dilution Measured Cardiac Output in the Critical Care Setting,” Critical Care Medicine, 13(4): 305, 1985. W. C. Shoemaker, P. L. Appel, H. B. Kram, R. C. Nathan, and J. L. Thompson, “Multicomponent Noninvasive Physiologic Monitoring of Circulatory Function,” Critical Care Medicine, 16(5): 482-490, 1988. N. J. Secher, A. Thomsen, and P. Arnsbo, “Measurement of Rapid Changes in Cardiac Stroke Volume,” An Evaluation of the Impedance Cardiography Method. Acta anaesth. scand., 21: 353-358, 1977. R. J. Demeter, S. A. Hanburger, P. D. Toth, C. T. Hawk, K. L. Parr, W. V. Judy, and M. E. Tavel, “Noninvasive Measurement of Systolic Time Intervals in the Dog Through the Use of Bioelectric Impedance,” Am. J. Noninvas. Cardiol., 2: 119-124, 1988. A. T. Edmunds, S. Godfrey, and M. Tooley, “Cardiac Output Measured by Transthoracic Impedance Cardiography at Rest, During Exercise and at Various Lung Volumes,” Clinical

[28]

[29]

[30]

[31]

[32]

[33]

[34]

201

Science, 63: 107-113, 1982. Y. Miyamoto, M. Takahashi, T. Tamura, T. Nakamura, T. Hiura, and M. Mikami, “Continuous Determination of Cardiac Output During Exercise by the Use of Impedance Plethysmography,” Med. & Biol. Eng. & comput., 19: 638-644, 1981. K. K. Teo, M. D. Hetherington, R. G. Haennel, P. V. Greenwood, R. E. Rossall, and T. Kappagoda, “Ardiac Output Measured by Impedance Cardiography During Maximal Exercise Tests,” Cardiovascular Research, 19: 737-743, 1985. Y. Zhang, M. Qu, J. G. Webster, W. J. Tompkins, B. A. Ward, and D. R. Bassett, “Cardiac Output Monitoring by Impedance Cardiography During Treadmill Exercise,” IEEE Transaction on Biomedical Engineering, 33(11):1037-1041, November, 1986. E. Enghoff, and O. Lheim, “A Comparison Between The Transthoracic Electrical Impedance Method And the Direct Fick And the Dye dilution Methods For Cardiac Output Measurements in Man,” Scand. J. clin. Lab. Invest., 39: 585-590, 1979. L. Y. Shyu, S. P. Reddy, J. H. Nagel, and N. Schneiderman, “New Signal Processing Techniques for Improved Reliability of Impedance Cardiography,” Proceedings of IEEE Engineering in Medicine & Biology Society 10th Annual International Conference, 41-42, 1988. J. H. Nagel, L. Y. Shyu, S. P. Reddy, B. E. Hurwitz, P. M. McCabe, and N. Schneiderman, “New Signal Processing Techniques For Improved Precision of Noninvasive Impedance Cardiography,” Annals of Biomedical Engineering., 17: 517-534, 1989. B. E. Hurwitz, L. Y. Shyu, S. P. Reddy, N. Schneiderman, and J. H. Nagel, “Coherent Ensemble Averaging Techniques For Impedance Cardiography,” Proceedings of the 3th annual IEEE symposium on computer-based medical systems, Chapel Hill, North Carolina, 228-235, June 3-6, 1990.