J Med Syst (2015) 39:35 DOI 10.1007/s10916-015-0223-5
MOBILE SYSTEMS
A Wearable Wireless ECG Monitoring System With Dynamic Transmission Power Control for Long-Term Homecare Yishan Wang & Sammy Doleschel & Ralf Wunderlich & Stefan Heinen
Received: 24 September 2014 / Accepted: 28 January 2015 # Springer Science+Business Media New York 2015
Abstract This paper presents a wearable wireless ECG monitoring system based on novel 3-Lead electrode placements for long-term homecare. The experiment for novel 3-Lead electrode placements is carried out, and the results show that the distance between limb electrodes can be significantly reduced. Based on the new electrode position, a small size sensor node, which is powered by a rechargeable battery, is designed to detect, amplify, filter and transmit the ECG signals. The coordinator receives the data and sends it to PC. Finally the signals are displayed on the GUI. In order to control the power consumption of sensor node, a dynamic power adjustment method is applied to automatically adjust the transmission power of the sensor node according to the received signal strength indicator (RSSI), which is related to the distance and obstacle between sensor node and coordinator. The system is evaluated when the user, who wears the sensor, is walking and running. A promising performance is achieved even under body motion. The power consumption can be significantly reduced with this dynamic power adjustment method. Keywords Wearable . Wireless . ECG monitor system . Novel 3-lead electrode placements . Dynamic power adjustment
Introduction As patients’ quality of care gains more and more attention, lots of noninvasive disease diagnosis methods are proposed [1, 2]. This article is part of the Topical Collection on Mobile Systems Y. Wang (*) : S. Doleschel : R. Wunderlich : S. Heinen Chair of Integrated Analog Circuits and RF Systems, RWTH Aachen University, D-52062 Aachen, Germany e-mail: [email protected]
Electrocardiogram (ECG) is one of the most important indicators for diagnosing many cardiac diseases. By measuring and amplifying body surface potentials at electrodes, ECG system presents potential differences at time series across these electrode placements [3]. In order to detect cardiac diseases earlier and reduce the hospitalization demands, the demand for continuous and real time monitoring using a body wearable wireless system is increasing. In the last century, many electrode placements were proposed, and the trend became more and more convenience. The first limb leads were defined by Einthoven in 1908 [4] (Fig. 1a). With these 3 electrodes on limb, 3 leads ECG can be detected. In 1944, Wilson proposed 6 precordial leads with 6 electrodes on chest [5] (Fig. 1b). Till this time, 12 leads ECG system was established. So as to provide more physical mobility for the patient, Mason and Likar published a new 12lead ECG system in 1966 to detect the 12 leads signals during exercise [6] (Fig. 1c). But this system still needs 10 electrodes on body. For the purpose of simplified lead system, in 1988, Dower developed a EASI system which utilized 5 electrodes to detect 3 EASI signals [7] (Fig. 1d). The 12-lead ECG signals can be reconstructed from these 3 EASI signals. However, with the electrode placements aforementioned, wires are still essential to connect the electrodes with the sensor. Therefore, the traditional electrode placements can not satisfy the wearable wireless ECG system in demand nowadays. In this century, several completely wireless ECG systems were published. Cao developed a three-pad ECG system. Three sensor nodes were worn around the heart to detect the signals from 3 dimensions, which were transmitted respectively to PC applying ZigBee [3] (Fig. 1e). 12-lead signals were reconstructed from the 3 dimension signals. This system is completely wireless. However, lots of efforts need to be paid on the time synchronization. Single-pad ECG systems were also published by [8–10] (Fig. 1f). But they can only detect one lead ECG signal. There is still no consensus on electrode
35
J Med Syst (2015) 39:35
Page 2 of 10
Fig. 1 The development of the lead system. (a) Limb leads [33] (b) Precodial leads [34] (c) Mason-Likar leads [35] (d) EASI system [35] (e) Three pad system [3] (f) Single pad system [10]
(a) Limb leads` [12]
(c) Mason-Likarr leads [14]
(e) ( Three pad system [3]
positions of an ECG system, which largely depend on a particular application [11]. In this paper, an experiment is designed to research the best electrode positions around heart. For the purpose of optimal quantity ECG signals and enough information, 3 lead signals are detected in this work. For a continuous, wireless and real time monitoring purpose, a small size and wireless ECG sensor needs to be designed. Due to the rapid development in telecommunication and information technologies, the wireless ECG system is burgeoning in recent years. However, to enable widespread use of such system, they must be very comfortable to wear, maintain the signal quality and transmit the data to hospital or
(b) Precoddial leads [13]
(d) EAS SI system [14]
(f) Single pad system [10]
doctors at real time. This implies that there are a lot of challenges on low noise, low power and comfortable ECG frontend design [12]. These techniques can allow patients to make the hospitalization at home and improve their life quality [13]. Based on the placements of the new electrodes obtained from the experiment, a wireless ECG sensor node with very small size is designed in this paper. A ZigBee coordinator and a GUI Data Acquisition Host are also built. These three parts constitute a compact, wearable, and wireless ECG system for continuous monitoring during daily activities. As a wireless body sensor, battery life time is always an important specification. For the purpose of long term homecare, a series of power mode control techniques is
J Med Syst (2015) 39:35
Page 3 of 10 35
applied in this work. Transmission is the most energy consuming operation in a sensor node. The radio transceiver’s transmission strength depends on its transmission power. Although a high transmission power level provides a good link quality, it raises the power consumption in sensor node. On the other side, low transmission power can degrade the energy consumption with the expense of diminished link quality. Accordingly, a tradeoff between link reliability and energy consumption must be obtained [14]. This paper establishes a dynamic adjustment rule for transmission power base on a sequence of measurements between received signal strength indicator (RSSI) and power levels. The rest of this paper is organized as follows. Section II describes the experiment and results of electrode placements. The design of the whole monitoring system is introduced in Section III, while the implementation is presented in section IV. Section V discusses the measurement results of dynamic transmission power control. The discussion and conclusion are presented in section VI and section VII respectively.
better to respectively keep on the either side of middle line, and that the signals slightly depended on the distance between them. The third placement was chosen as the best placement for new right and left arm electrodes, which is also applied in this paper. The study in [15] also analyzed the placements for left leg electrode. But only 5 placements were tested. The result showed that it was better to keep the left leg electrode on the left side of the middle line. As a result, this paper employs 9 new placements of left leg electrode in both horizontal and vertical direction (as shown in Fig. 3). With the same method in [15], three lead signals (Lead I, Lead II and Lead III) of these new placements are measured and compared with the signals of Mason-Likar limb lead system, which is used as a standard lead system. The correlation coefficients between them are analyzed. With the purpose of enhancing comparability, the measurements for standard and new placements are implemented on the same volunteer and at the same time. The electrodes used in the experiments are supplied by Ambu [16]. The signals are collected by the wireless ECG system published in [17].
Method and system (b) Results Novel 3-Lead electrode placements In order to make the system compact and wearable, new electrode positions need to be determined. Based on our previous research in [15], a new experiment is designed to confirm the new placements for three limb electrodes, which have significantly short distance between each other. (a) Experiment The experiments in [15] tested 10 placements for right arm, left arm and left leg electrodes. Mason-Likar limb electrode placement [6] which was popularly used in the last four decades was considered as standard lead system. The results of the experiments for right arm and left arm electrodes (as shown in Fig. 2) indicated that these two electrodes were
(a) The first
(b) The second
According to the experiment for the placements of left leg electrode, three lead signals of standard and new lead system are measured and compared. The correlation coefficients between them are calculated (as listed in Table 1). It can be obviously seen that, the results highly depend on the horizontal direction, but slightly on the vertical direction. Three lead signals of LL3 location in different vertical distances are displayed in Fig. 4. Compared to the standard lead system, the signals only have the differences in amplitude. The experiment of the novel electrode placements convincingly evidences that, the distance between electrodes can be significantly reduced. In order to satisfy both the requirements of small size and high correlation coefficient with standard lead, the placement with 5 cm LL3 is selected as the best new left leg electrode location.
(c) The third
(d) The fourth
(e) The fifth
Fig. 2 The placements of right arm and left arm electrodes in [15]. (a) The first. (b) The second. (c) The third. (d) The fourth. (e) The fifth
35
J Med Syst (2015) 39:35
Page 4 of 10
Fig. 3 The new placements of this paper for left leg electrode (a) 4 cm in vertical (b) 5 cm in vertical (c) 6 cm in vertical
(a) 4 cm in vertical Sensor node design The considered wireless ECG monitoring system constitutes a sensor node, a coordinator and a GUI. The signals detected by the sensor are transmitted to the coordinator and finally displayed and analyzed on PC. The diagram of the whole system is shown in Fig. 5. According to the results of the electrode placements, t-he final arrangement of the electrodes and the size of sensor node are shown in Fig. 6. The size of the sensor node can be controlled in 5.5 cm×2.5 cm. The sensor node includes the analog front end which amplifies and filters the signal, the microcontroller and RF transceiver that samples and transmits the signals. The whole sensor node is powered by a rechargeable 600 mAh battery. The recharging circuit is also designed in the sensor node.
Analog front end is the core for the performance of the ECG system. The common mode noise, especially the 50/ 60 Hz electromagnetic interference from the power line Table 1 The correlation coefficient of three lead signals between standard and new placements for left leg electrode
4 cm
5 cm
6 cm
Standard placement
LL1 LL2 LL3 LL1 LL2 LL3 LL1 LL2 LL3
(c) 6 cm in vertical
should be rejected in this stage. Besides, the dc offset voltage caused by the body motion and electrode mismatch need to be removed before the instrumentation amplifier (INA). This paper applies the front end circuit proposed in our previous work [17], which includes ac coupling circuit, instrumentation amplifier and bandpass filter. The dc offset voltage caused by body motion is reduced by ac coupling circuit, while the 50/ 60 Hz electromagnetic interference is reduced by the high common mode rejection ratio (CMRR) instrumentation amplifier. The results in [17] perform improved cancellation of body motion effect and high suppression of common mode noise. The front end can detect two lead signals (Lead II and Lead III). The Lead I signal can be calculated as Lead I ¼ Lead II−Lead III
ð1Þ
(b) Microcontroller and RF transceiver
(a) Analog front end
New placement
(b) 5 cm in vertical
Lead I
Lead II
Lead III
0.91 0.88 0.87 0.91 0.90 0.91 0.91 0.91 0:90
0.34 0.61 0.86 0.46 0.69 0.91 0.60 0.83 0.93
-0.29 0.26 0.88 -0.34 0.45 0.93 -0.19 0.70 0.96
TI CC2530 System-on-Chip with ZigBee protocol [18], which features very low power consumption due to its fast speed from sleep to active mode, is applied in this work. It includes the microcontroller which samples the signals from front end with 200 Hz sampling frequency, and the transceiver which transmits the data to coordinator. A 50 Ω Inverted F PCB antenna (IFA) [19] is also designed. In order to reduce the power consumption, the data is sent every 100 ms with 20 samples in every packet. The microcontroller is running in low power mode. During the absence of sampling and transmitting task, it switches to sleep mode. And it can be woken up by the ADC interruption in 4 μs with 0.2 mA current consumption. If there is no connection to the coordinator, it will enter into the deep sleep mode which consumes only 1 μA current. (c) Coordinator The Coordinator consists of USB stick for debugging, USB SPI interface (FT220X) [20], system on chip CC2530 [18]
J Med Syst (2015) 39:35
Page 5 of 10 35
4cm−−LeadI
4cm−−LeadII
1
1 Standard Lead System New Lead System
0.5
0
1.5
2
2.5 Time (s)
3
3.5
0.5
0
−0.5 1
4
1.5
2
5cm−−LeadI
2.5 Time (s)
0.5
0
1.5
2
2.5 Time (s)
3
3.5
−0.5 1
4
1.5
2
2.5 Time (s)
3
3.5
−0.5 1
4
1.5
2
3
3.5
4
2.5 Time (s)
3
3.5
4
6cm−−LeadIII
Voltage (V)
Standard Lead System New Lead System
0.5
0
−0.5 1
4
1 Standard Lead System New Lead System
Voltage (V) 2.5 Time (s)
3.5
0
6cm−−LeadII
0
3
0.5
1
0.5
2.5 Time (s)
Standard Lead System New Lead System
0
Standard Lead System New Lead System
2
2
5cm−−LeadIII
0.5
6cm−−LeadI
1.5
1.5
1
Voltage (V)
Voltage (V)
Voltage (V)
−0.5 1
4
Standard Lead System New Lead System
1
Voltage (V)
3.5
0
1 Standard Lead System New Lead System
−0.5 1
3
0.5
5cm−−LeadII
1
−0.5 1
Standard Lead System New Lead System Voltage (V)
Voltage (V)
Voltage (V)
Standard Lead System New Lead System
−0.5 1
4cm−−LeadIII
1
1.5
2
2.5 Time (s)
3
3.5
4
0.5
0
−0.5 1
1.5
2
2.5 Time (s)
3
3.5
4
Fig. 4 The comparison of signals between standard placement and new placements
and 50 Ω Inverted F PCB antenna (IFA). It receives the data from sensor node and sends back the acknowledgment message with Tx output power adjustment suggestion. The received data is transmitted to FT220X and then sent to PC. The detailed process of the communication between sensor node and coordinator is shown in Fig. 7.
For the purpose of low power consumption of the transceiver, this work employed a dynamic power adjustment method [21]. Every data packet the coordinator receives includes the RSSI. RSSI, which ranges from 0 dBm to -100 dBm, is computed internally in the radio by averaging the signal power over eight symbol periods of the incoming packet. It is related to the link state (distance and obstacle) between sensor node and coordinator. Therefore, it is a good indicator to adjust the Tx output power of the sensor node. With the purpose to find a suitable method to adjust the Tx output power, an experiment is designed. Because distance is an important factor to affect the RSSI value, the Tx output power is recorded at every distance, as shown in Fig. 8. In order to get the same signal strength at every distance, the output power is adjusted until the RSSI reaches between -60 dBm and -70 dBm. To avoid other factors affecting RSSI, the measurement is done in an open area. The result indicates that the Tx output power does not change linearly, but changes faster in lower levels than in higher levels. The relationship between Tx output power and current consumption of transceiver is listed in Table 2. It states that it can
Dynamic transmission power control In the most of the body sensor network system, the transceiver always consumes most of the power of sensor node. In the body sensor application, the patient who is carrying the sensor will move in a certain range. To successfully perform over a long term and monitor signal continuously, the transmitter power should not be set in a fixed level, but be adjustable according to the received signal strength. At night, the patient would stay at the same place. If the coordinator is put close to the sensor, and the Tx output power of the sensor is adjusted to a lower level, the power consumption can be significantly reduced. Fig. 5 System diagram
RA INA
BPF
LA
A INA
LL
BPF
D
MCU
RX TX
RX TX
MCU
35
J Med Syst (2015) 39:35
Page 6 of 10 5
Tx Output Power (dBm)
0 −5 −10 −15 −20
Fig. 6 The design of the sensor node −25
save 11 mA in lowest level compared with the highest level. But at the lower level, the current consumption does not change a lot. According to these two results, the Tx output power is set at 9 levels (as listed in Table 3) to achieve fast adjustable performance. It adjusts fast at low levels but slowly at high levels. This setting can largely avoid the link break between sensor node and coordinator, when the link state varies quickly. The previous measurement of this work also shows that when the RSSI becomes lower than -90 dBm, the data can not be received successfully. As a result, the rules of the adjustment are set as: RSSI ≤ −80dBm : PowerLevel ¼ 8 −80dBm < RSSI < −75dBm : PowerLevel þ 1 RSSI > −65dBm : PowerLevel−1
Wake up every 1s
No
ð2Þ
0
5 10 15 20 Distance between sensor node and coordinator (m)
Fig. 8 Transmission power changes according to distance
With these rules, the RSSI is controlled between -65 dBm and -75 dBm. The adjustment suggestion will be sent with the acknowledgment to the sensor node. The detailed information of the communication is shown in Fig. 7.
GUI data acquisition host A GUI Data Acquisition Host is built to set the UART parameters and display the signals received from the Coordinator. The QRS complexes can also be detected at real time using the Hamilton-Tompkins algorithm [22], and the heart rate is calculated according to the period between QRS complexes.
Sensor Node
Coordinator
Initialization Search coordinator
Initialization Wait for connection
Find coordinator? Match network key? Yes
Enter into deep sleep mode (Power Mode 2)
Send connection ask Wait for answer No
No
Yes Connection allowed Send response
Succesful? Yes
Wait for data
Enable ADC Interruption ADC interruption, Wake up No Enter into sleep mode (Power Mode 1)
Sending task? Yes Send the data Wait for answer
No Yes
3 times?
No
Receive succesfully?
Read RSSI Send acknowledgement with Tx power adjustment suggestion
Yes
Fig. 7 Flow chart of communication between sensor node and coordinator
No
Yes
Send succesfully?
Adjust the Tx power
25
Send data to PC
J Med Syst (2015) 39:35 Table 2
Page 7 of 10 35
Current consumption of CC2530 at every Tx output power [18]
Tx output power (dBm)
4.5
2.5
1
-0.5
-1.5
-3
-4
-6
-8
-10
-12
-14
-16
-18
-20
-22
Current consumption (mA)
34
31
29
28
27
27
26
26
25
25
25
25
25
24
24
23
Implementation and measurement results
Dynamic transmission power control
ECG monitoring system
The best way to evaluate the performance of dynamic transmission power control is to do the measurement under body motion. As the distance is the main factor affecting the adjustment, the measurement is designed as follows: the volunteer, who wears the sensor node, starts walking away from the coordinator, and then turns back at 20 m and walks towards to the coordinator with different speeds. The fluctuations of Tx output power level and RSSI value during this procedure are recorded (as shown in Fig. 12). The results illustrate that, no matter the volunteer walks fast or slow, the power level can be adjusted automatically and the RSSI is kept around -70 dBm. During walking under normal speed, about 20 % power is saved in 45 s. As exhibited in Fig. 12d, when the volunteer is under rest, the power level and RSSI can be kept stable and 30 % power is saved. This strongly demonstrates that, with the dynamic transmission power control, the power consumption can be significantly reduced when the patient is sleeping during night or remaining motionless. Figure 12 also indicates that when the volunteer is walking away and towards to coordinator, the power level does not change linearly, but with some fluctuations. This phenomenon is caused by the surrounding environment. The power level is adjusted according to the RSSI. But the RSSI not only depends on the distance, but also on the surrounding environment. The walls and other obstacles will also affect the RSSI. Consequently, when there are some obstacles between the
Figure 9a shows the photograph of sensor node, the size of which is controlled in 5.5 cm×2.5 cm. Bottom view is the analog front end and the connectors to electrodes. Top view shows the rechargeable battery and the charging port. The transceiver and the antenna are implanted below the battery. Figure 9b shows the photograph of the coordinator. The maximum communication distance between sensor node and coordinator is 30 m, which can easily fulfill home usage applications. The current is measured when there is no connection with the coordinator. As described before, the sensor node enters into deep sleep mode during this situation and consumes about 5 mA current. After every 1 s it wakes up and searches the coordinator. The current consumption of this process is clearly shown in Fig. 10a. As shown in Fig. 10b, the current consumption of sensor node under communication with coordinator is measured. It can be apparently seen that, the data is sent every 100 ms. During every period, three current peaks, which denote sending data, receiving acknowledgment and sending successful feedback respectively, are achieved. The period between the ripples tangibly reveals that, the CC2530 enter into sleep mode during the task free time. Owing to the power mode control, the sensor node saves plenty of power, especially when there is no connection with the coordinator. Even if the sensor works continuously, the battery life can be more than 52 h. The GUI data acquisition host is shown in Fig. 11. It can successfully detect the QRS complexes and calculate the heart rate at real time. The sensitivity of QRS complex detection is 97.22 % when the body is peaceful and 91.25 % when the body is running. The GUI can also show an alarm when the heart rate is abnormal. The whole system with compact size and light weight establishes stable communication among sensor node, coordinator and GUI. Along with that, high quality signals are achieved even under body motion (walking and running). Table 3
Tx output power level settings
Power level
0
1
2
3
4
5
6 7
8
Tx output power (dBm) -22 -18 -8 -4 -1.5 -0.5 1 2.5 4.5
(a) Sensor node
(b) Coordinator
Fig. 9 Photograph of the system. (a) Sensor node (b) Coordinator
35
J Med Syst (2015) 39:35
Page 8 of 10 40
Current (mA)
30
20
10
0
1
2
3
4
5
6
Time (s)
(a) During coordinator absent 40
The fluctuations in low levels are stronger than in high levels, because the adjustment steps in low levels are bigger than in high level. It can be avoided by reducing the steps. According to the measurement in Fig. 12, the Tx output power varies very fast in lower distances. In consequence, small steps in low levels will cause the link disconnection when the link state varies quickly, especially when the sensor moves very fast away from the coordinator. If the Tx output power cannot be adjusted fast enough, the data cannot be received successfully. To keep the stability of the link, big steps in low power levels are still used in adjustment.
Current (mA)
30
20
Discussion and conclusion
10
Discussion 0 50
100
150
200
250 300 Time (ms)
350
400
450
(a) Novel limb electrode placements
(b) During communication with coordinator Fig. 10 Current consumption. (a) During coordinator absent (b) During communication with coordinator
sensor and coordinator, the power level will also increase, even though the distance is very short.
Fig. 11 GUI data acquisition host
Compare to the design in [23–26], the novel placements for limb electrodes presented in this paper is possible to connect the electrodes with sensor node completely wireless. This system has more flexibility and compactness for the patients than the design in [3, 27]. It only needs one sensor node in small size and avoids the time synchronization between every
J Med Syst (2015) 39:35
Page 9 of 10 35
−60
2
−70
−80
0
−80
−90 30
−2
−60
2
−70
0
20
25
0
5
10
15
(a) Fast walking
Tx Output Power Level
−60
2
−70
−80
0
−80
−90 60
−2
−60
2
−70
0
50
−40
4
4
40
−30
−50
−50
30 Time (s)
−90 45
6
6
20
40
Tx Power Level RSSI
8
−40
RSSI (dBm)
Tx Output Power Level
8
10
35
10
−30 Tx Power Level RSSI
0
30
(b) Normal walking
10
−2
20 25 Time (s)
0
5
(c) Slow walking
10
15
20 Time (s)
25
30
35
RSSI (dBm)
15 Time (s)
RSSI (dBm)
4
4
10
Tx Output Power Level
Tx Output Power Level
−50
−50
5
−40
6
6
0
−30 Tx Power Level RSSI
8
−40
RSSI (dBm)
Tx Power Level RSSI
8
−2
10
−30
10
−90 40
(d) Resting
Fig. 12 Power level adjustment and RSSI during body motion and resting. (a) Fast walking (b) Normal walking (c) Slow walking (d) Resting
sensor node. Although it is bigger in size than [9, 10, 13], the proposed system can detect more lead signals and supply more information for diagnosis.
defined. It can be adjusted much faster than the 32 levels in [21], and avoid the disconnection when the link state varies rapidly.
(b) Wireless ECG monitoring system Based on the new electrode placements, the size of the sensor is controlled in 5.5 cm×2.5 cm, which can satisfy the long term wearable requirement. Compared to the design in [26–28], this system uses simplified circuit, but can still get the high quality signals and meanwhile make the size of the sensor compact. Unlike the design in [13], this design can monitor the ECG signals continuously. Without the hand held device (HHD) which is designed in [29], the system becomes much more flexible. (c) Dynamic transmission power control Many transmission power control protocols have been employed in wireless sensor networks [30]. However most of these studies have been targeted for static networks [31, 32]. The studies [14] and [21] published their researches on transmission power control in body area network. Compare to [14], the proposed work simplified the adjustment protocol according to our practical application. Based on the measurement of the Tx output power at every distance and the current consumption at every Tx output power, 9 power levels are
Conclusion This paper introduces a wireless and wearable ECG monitoring system for long term homecare. The novel electrode placements account for the possiblity of compact size and long term wearable sensor design. It can be successfully applied in the wireless and wearable ECG sensor system. The sensor node is implemented in small size and with power control. The communication between the sensor node and coordinator performs securely and steadily. The power mode control on sensor node substantially reduces the power consumption during the task absence time. It tremendously extends the life time of battery. The dynamic transmission power control is employed. The Tx output power of the transceiver on sensor node is automatically adjusted according to the RSSI value. This method is tested under both the fast and slow body motion. The results demonstrate that, the adjustment works agilely, accurately and stably. It reduces 20 % power consumption during normal walking and 30 % during resting. Obviously, the dynamic transmission power control will contribute a lot in extending the battery life time. Overall, the whole system can certainly
35
Page 10 of 10
satisfy the long term homecare usage and provide more physical mobility to patients.
References 1. Wang, Y., Yu, K., Wang, D., Zhao, C., Wang, L. and Wang, P., BMulti-Model Diagnosis Method for Lung Cancer based on MOSSAW Breath Detecting e-Nose,^ in Proceedings of The 14th International Symposium on Olfaction and Electronic Nose, 2011. 2. Wang, Y., Hu, Y., Wang, D., Yu, K., Wang, L., Zou, Y., Zhao, C., Zhang, X., Wang, P., and Ying, K., The analysis of volatile organic compounds biomarkers for lung cancer in exhaled breath, tissues and cell lines. Cancer. Biomark. 11:129–137, 2012. 3. Cao, H., Li, H., Stocco, L., and Leung, V. C. M., Wireless three-pad ECG system: challenges, design, and evaluations. Commun. Net. J. 13:113–124, 2011. 4. Einthoven, W., Weiteres ueber das elektrokardiogram. Pflueger Arch. ges. Physiol. 122:517–584, 1908. 5. Wilson, F. N., Johnston, F. D., Rosenbaum, F. F., Erlanger, H., Kossmann, C. E., Hecht, H., Cotrim, N., Oliveira, R. M. d., Scarsi, R. and Barker, P. S., BThe precordial electrocardiogram,^ AHJ, vol. 27, pp. 19-85, 1944. 6. Mason, R. E., and Likar, I., A new system of multiple-lead exercise electrocardiography. AHJ. 71:196–205, 1966. 7. Dower, G. E., Yakush, A., Nazzal, S. B., Jutzy, R. V., and Ruiz, C. E., Deriving the 12-lead electrocardiogram from four (EASI) electrodes. J. Electrocardiol. 21:182–187, 1988. 8. Munshi, M., Xu, X., Zou, X., Soetiono, E., Teo, C. S. and Lian, Y., BWireless ECG plaster for body sensor network,^ in Proceedings of ISSS-MDBS 2008, 2008. 9. Masse, F., Penders, J., Serteyn, A., Bussel, M. van and Arends, J., BMiniaturized wireless ECG-monitor for real-time detection of epileptic seizures,^ in Proceedings of Wireless Health 2010, 2010. 10. Altini, M., Polito, S., Penders, J., Kim, H., Helleputte, N. Van, Kim, S. and Yazicioglu, F., BAn ECG patch combining a customized ultralow-power ECG SoC with bluetooth low energy for long term ambulatory monitoring,^ in Proceedings of the 2nd Conference on Wireless Health, 2011. 11. Hoekema, R., Uijen, G., and van Oosterom, A., On selecting a body surface mapping procedure. J. Electrocardiol. 32:93–101, 1999. 12. Torfs, Yazicioglu, T. R., Kim, S., Kim, H., Hoof, C. Van, Buxi, D., Romero, I. a. W. J., Masse, F. and Penders, J., BUltra low power wireless ECG system with beat detection and real time impedance measurement,^ in Proceedings of IEEE BioCAS 2010 Conference, 2010. 13. Gomez-Clapers, J., and Casanella, R., A fast and easy-to-use ECG acquisition and heart rate monitoring system using a wireless steering wheel. Sens. J. IEEE. 12:610–616, 2012. 14. Kim, S., and Eom, D., Link-state-estimation-based transmission power control in wireless body area networks. IEEE J-BHI. 18: 1294–1302, 2014. 15. Wang, Y., Wunderlich, R. and Heinen, S., BDesign and Evaluation of a Novel Wireless Reconstructed 3-Lead ECG Monitoring System,^ in Proceedings of IEEE BioCAS 2013 Conference, 2013. 16. Ambu® BlueSensor M. Ambu Inc., 2013.
J Med Syst (2015) 39:35 17. Wang. Y., Wunderlich. R. and Heinen. S., BA low noise wearable wireless ECG system with body motion cancellation for long term homecare,^ in Proceedings of IEEE Healthcom 2013 Conference, 2013. 18. A true system-on-chip solution for 2.4-GHz IEEE 802.15.4 and ZigBee applications. Texas Instruments Inc., 2011. 19. Andersen, A., 2.4 GHz inverted F antenna. Texas Instruments Inc., 2008. 20. USB 4-BIT SPI/FT1248 IC FT220X. Future Technology Devices International Ltd., 2014. 21. Xiao, S., Dhamdhere, A., Sivaraman, V., and Burdett, A., Transmission power control in body area sensor networks for healthcare monitoring. IEEE J-SAC. 27:37–48, 2009. 22. Pan, J., and Tompkins, W. J., A real-time QRS detection algorithm. IEEE Trans. Biomed. Eng. BME-32:230–236, 1985. 23. Burns, A., Doheny, E., Greene, B., Foran, T., Leahy, D., O’Donovan, K. and McGrath, M., BSHIMMER™: an extensible platform for physiological signal capture,^ in Proceedings of Annual International Conference of IEEE EMBC 2010, 2010. 24. Gaxiola-Sosa, J., Mohsin, N., Palliyali, A., Tafreshi, R. and Entesari, K., BA portable 12-lead ECG wireless medical system for continuous cardiac-activity monitoring,^ in Proceedings of MECBME 2014, 2014. 25. Gao, H., Duan, X., Guo, X., Huang, A. and Jiao, B., BDesign and tests of a smartphones-based multi-lead ECG monitoring system,^ in Proceedings of Annual International Conference of IEEE EMBC 2013, 2013. 26. Tan, T.-H., Chang, C.-S., Huang, Y.-F., Chen, Y.-F., and Lee, C., Development of a portable linux-based ECG measurement and monitoring system. J. Med. Syst. 35:559–569, 2011. 27. Winokur, E., Delano, M., and Sodini, C., A wearable cardiac monitor for long-term data acquisition and analysis. IEEE Trans. Biomed. Eng. 60:189–192, 2013. 28. Chen, H., Wu, W., and Lee, J., A WBAN-based real-time electroencephalogram monitoring system: design and implementation. J. Med. Syst. 34:303–311, 2010. 29. Fensli, R., Dale, J., O’Reilly, P., O’Donoghue, J., Sammon, D., and Gundersen, T., Towards improved healthcare performance: examining technological possibilities and patient satisfaction with wireless body area networks. J. Med. Syst. 34:767–775, 2010. 30. Pantazis, N., and Vergados, D., A survey on power control issues in wireless sensor networks. IEEE Commun. Surv. Tutor. 9:86–107, 2007. 31. Zhou, G., He, T., Krishnamurthy, S. and Stankovic, J. A., BImpact of radio irregularity on wireless sensor networks,^ in Proceedings of MobiSys 2004, 2004. 32. Zhao, J. and Govindan, R., BUnderstanding packet delivery performance in dense wireless sensor networks,^ in Proceedings of the 1st International Conference on Embedded Networked Sensor Systems, 2003. 33. Malmivuo, J., and Plonsey, R., Bioelectromagnetism: principles and applications of bioelectric and biomagnetic fields. Oxford University Press, New York, 1995. 34. Yanowitz, F. G., Introduction to ECG interpretation V8.0. Eccles Health Sciences Library University of Utah, Salt Lake City, 2012. 35. Welinder, A., Sörnmo, L., Field, D. Q., Feldman, C. L., Pettersson, J., Wagner, G. S., and Pahlm, O., Comparison of signal quality between easy and Mason-Likar 12-Lead electrocardiograms during physical activity. AJCC. 13:228–234, 2004.
MOBILE SYSTEMS
A Wearable Wireless ECG Monitoring System With Dynamic Transmission Power Control for Long-Term Homecare Yishan Wang & Sammy Doleschel & Ralf Wunderlich & Stefan Heinen
Received: 24 September 2014 / Accepted: 28 January 2015 # Springer Science+Business Media New York 2015
Abstract This paper presents a wearable wireless ECG monitoring system based on novel 3-Lead electrode placements for long-term homecare. The experiment for novel 3-Lead electrode placements is carried out, and the results show that the distance between limb electrodes can be significantly reduced. Based on the new electrode position, a small size sensor node, which is powered by a rechargeable battery, is designed to detect, amplify, filter and transmit the ECG signals. The coordinator receives the data and sends it to PC. Finally the signals are displayed on the GUI. In order to control the power consumption of sensor node, a dynamic power adjustment method is applied to automatically adjust the transmission power of the sensor node according to the received signal strength indicator (RSSI), which is related to the distance and obstacle between sensor node and coordinator. The system is evaluated when the user, who wears the sensor, is walking and running. A promising performance is achieved even under body motion. The power consumption can be significantly reduced with this dynamic power adjustment method. Keywords Wearable . Wireless . ECG monitor system . Novel 3-lead electrode placements . Dynamic power adjustment
Introduction As patients’ quality of care gains more and more attention, lots of noninvasive disease diagnosis methods are proposed [1, 2]. This article is part of the Topical Collection on Mobile Systems Y. Wang (*) : S. Doleschel : R. Wunderlich : S. Heinen Chair of Integrated Analog Circuits and RF Systems, RWTH Aachen University, D-52062 Aachen, Germany e-mail: [email protected]
Electrocardiogram (ECG) is one of the most important indicators for diagnosing many cardiac diseases. By measuring and amplifying body surface potentials at electrodes, ECG system presents potential differences at time series across these electrode placements [3]. In order to detect cardiac diseases earlier and reduce the hospitalization demands, the demand for continuous and real time monitoring using a body wearable wireless system is increasing. In the last century, many electrode placements were proposed, and the trend became more and more convenience. The first limb leads were defined by Einthoven in 1908 [4] (Fig. 1a). With these 3 electrodes on limb, 3 leads ECG can be detected. In 1944, Wilson proposed 6 precordial leads with 6 electrodes on chest [5] (Fig. 1b). Till this time, 12 leads ECG system was established. So as to provide more physical mobility for the patient, Mason and Likar published a new 12lead ECG system in 1966 to detect the 12 leads signals during exercise [6] (Fig. 1c). But this system still needs 10 electrodes on body. For the purpose of simplified lead system, in 1988, Dower developed a EASI system which utilized 5 electrodes to detect 3 EASI signals [7] (Fig. 1d). The 12-lead ECG signals can be reconstructed from these 3 EASI signals. However, with the electrode placements aforementioned, wires are still essential to connect the electrodes with the sensor. Therefore, the traditional electrode placements can not satisfy the wearable wireless ECG system in demand nowadays. In this century, several completely wireless ECG systems were published. Cao developed a three-pad ECG system. Three sensor nodes were worn around the heart to detect the signals from 3 dimensions, which were transmitted respectively to PC applying ZigBee [3] (Fig. 1e). 12-lead signals were reconstructed from the 3 dimension signals. This system is completely wireless. However, lots of efforts need to be paid on the time synchronization. Single-pad ECG systems were also published by [8–10] (Fig. 1f). But they can only detect one lead ECG signal. There is still no consensus on electrode
35
J Med Syst (2015) 39:35
Page 2 of 10
Fig. 1 The development of the lead system. (a) Limb leads [33] (b) Precodial leads [34] (c) Mason-Likar leads [35] (d) EASI system [35] (e) Three pad system [3] (f) Single pad system [10]
(a) Limb leads` [12]
(c) Mason-Likarr leads [14]
(e) ( Three pad system [3]
positions of an ECG system, which largely depend on a particular application [11]. In this paper, an experiment is designed to research the best electrode positions around heart. For the purpose of optimal quantity ECG signals and enough information, 3 lead signals are detected in this work. For a continuous, wireless and real time monitoring purpose, a small size and wireless ECG sensor needs to be designed. Due to the rapid development in telecommunication and information technologies, the wireless ECG system is burgeoning in recent years. However, to enable widespread use of such system, they must be very comfortable to wear, maintain the signal quality and transmit the data to hospital or
(b) Precoddial leads [13]
(d) EAS SI system [14]
(f) Single pad system [10]
doctors at real time. This implies that there are a lot of challenges on low noise, low power and comfortable ECG frontend design [12]. These techniques can allow patients to make the hospitalization at home and improve their life quality [13]. Based on the placements of the new electrodes obtained from the experiment, a wireless ECG sensor node with very small size is designed in this paper. A ZigBee coordinator and a GUI Data Acquisition Host are also built. These three parts constitute a compact, wearable, and wireless ECG system for continuous monitoring during daily activities. As a wireless body sensor, battery life time is always an important specification. For the purpose of long term homecare, a series of power mode control techniques is
J Med Syst (2015) 39:35
Page 3 of 10 35
applied in this work. Transmission is the most energy consuming operation in a sensor node. The radio transceiver’s transmission strength depends on its transmission power. Although a high transmission power level provides a good link quality, it raises the power consumption in sensor node. On the other side, low transmission power can degrade the energy consumption with the expense of diminished link quality. Accordingly, a tradeoff between link reliability and energy consumption must be obtained [14]. This paper establishes a dynamic adjustment rule for transmission power base on a sequence of measurements between received signal strength indicator (RSSI) and power levels. The rest of this paper is organized as follows. Section II describes the experiment and results of electrode placements. The design of the whole monitoring system is introduced in Section III, while the implementation is presented in section IV. Section V discusses the measurement results of dynamic transmission power control. The discussion and conclusion are presented in section VI and section VII respectively.
better to respectively keep on the either side of middle line, and that the signals slightly depended on the distance between them. The third placement was chosen as the best placement for new right and left arm electrodes, which is also applied in this paper. The study in [15] also analyzed the placements for left leg electrode. But only 5 placements were tested. The result showed that it was better to keep the left leg electrode on the left side of the middle line. As a result, this paper employs 9 new placements of left leg electrode in both horizontal and vertical direction (as shown in Fig. 3). With the same method in [15], three lead signals (Lead I, Lead II and Lead III) of these new placements are measured and compared with the signals of Mason-Likar limb lead system, which is used as a standard lead system. The correlation coefficients between them are analyzed. With the purpose of enhancing comparability, the measurements for standard and new placements are implemented on the same volunteer and at the same time. The electrodes used in the experiments are supplied by Ambu [16]. The signals are collected by the wireless ECG system published in [17].
Method and system (b) Results Novel 3-Lead electrode placements In order to make the system compact and wearable, new electrode positions need to be determined. Based on our previous research in [15], a new experiment is designed to confirm the new placements for three limb electrodes, which have significantly short distance between each other. (a) Experiment The experiments in [15] tested 10 placements for right arm, left arm and left leg electrodes. Mason-Likar limb electrode placement [6] which was popularly used in the last four decades was considered as standard lead system. The results of the experiments for right arm and left arm electrodes (as shown in Fig. 2) indicated that these two electrodes were
(a) The first
(b) The second
According to the experiment for the placements of left leg electrode, three lead signals of standard and new lead system are measured and compared. The correlation coefficients between them are calculated (as listed in Table 1). It can be obviously seen that, the results highly depend on the horizontal direction, but slightly on the vertical direction. Three lead signals of LL3 location in different vertical distances are displayed in Fig. 4. Compared to the standard lead system, the signals only have the differences in amplitude. The experiment of the novel electrode placements convincingly evidences that, the distance between electrodes can be significantly reduced. In order to satisfy both the requirements of small size and high correlation coefficient with standard lead, the placement with 5 cm LL3 is selected as the best new left leg electrode location.
(c) The third
(d) The fourth
(e) The fifth
Fig. 2 The placements of right arm and left arm electrodes in [15]. (a) The first. (b) The second. (c) The third. (d) The fourth. (e) The fifth
35
J Med Syst (2015) 39:35
Page 4 of 10
Fig. 3 The new placements of this paper for left leg electrode (a) 4 cm in vertical (b) 5 cm in vertical (c) 6 cm in vertical
(a) 4 cm in vertical Sensor node design The considered wireless ECG monitoring system constitutes a sensor node, a coordinator and a GUI. The signals detected by the sensor are transmitted to the coordinator and finally displayed and analyzed on PC. The diagram of the whole system is shown in Fig. 5. According to the results of the electrode placements, t-he final arrangement of the electrodes and the size of sensor node are shown in Fig. 6. The size of the sensor node can be controlled in 5.5 cm×2.5 cm. The sensor node includes the analog front end which amplifies and filters the signal, the microcontroller and RF transceiver that samples and transmits the signals. The whole sensor node is powered by a rechargeable 600 mAh battery. The recharging circuit is also designed in the sensor node.
Analog front end is the core for the performance of the ECG system. The common mode noise, especially the 50/ 60 Hz electromagnetic interference from the power line Table 1 The correlation coefficient of three lead signals between standard and new placements for left leg electrode
4 cm
5 cm
6 cm
Standard placement
LL1 LL2 LL3 LL1 LL2 LL3 LL1 LL2 LL3
(c) 6 cm in vertical
should be rejected in this stage. Besides, the dc offset voltage caused by the body motion and electrode mismatch need to be removed before the instrumentation amplifier (INA). This paper applies the front end circuit proposed in our previous work [17], which includes ac coupling circuit, instrumentation amplifier and bandpass filter. The dc offset voltage caused by body motion is reduced by ac coupling circuit, while the 50/ 60 Hz electromagnetic interference is reduced by the high common mode rejection ratio (CMRR) instrumentation amplifier. The results in [17] perform improved cancellation of body motion effect and high suppression of common mode noise. The front end can detect two lead signals (Lead II and Lead III). The Lead I signal can be calculated as Lead I ¼ Lead II−Lead III
ð1Þ
(b) Microcontroller and RF transceiver
(a) Analog front end
New placement
(b) 5 cm in vertical
Lead I
Lead II
Lead III
0.91 0.88 0.87 0.91 0.90 0.91 0.91 0.91 0:90
0.34 0.61 0.86 0.46 0.69 0.91 0.60 0.83 0.93
-0.29 0.26 0.88 -0.34 0.45 0.93 -0.19 0.70 0.96
TI CC2530 System-on-Chip with ZigBee protocol [18], which features very low power consumption due to its fast speed from sleep to active mode, is applied in this work. It includes the microcontroller which samples the signals from front end with 200 Hz sampling frequency, and the transceiver which transmits the data to coordinator. A 50 Ω Inverted F PCB antenna (IFA) [19] is also designed. In order to reduce the power consumption, the data is sent every 100 ms with 20 samples in every packet. The microcontroller is running in low power mode. During the absence of sampling and transmitting task, it switches to sleep mode. And it can be woken up by the ADC interruption in 4 μs with 0.2 mA current consumption. If there is no connection to the coordinator, it will enter into the deep sleep mode which consumes only 1 μA current. (c) Coordinator The Coordinator consists of USB stick for debugging, USB SPI interface (FT220X) [20], system on chip CC2530 [18]
J Med Syst (2015) 39:35
Page 5 of 10 35
4cm−−LeadI
4cm−−LeadII
1
1 Standard Lead System New Lead System
0.5
0
1.5
2
2.5 Time (s)
3
3.5
0.5
0
−0.5 1
4
1.5
2
5cm−−LeadI
2.5 Time (s)
0.5
0
1.5
2
2.5 Time (s)
3
3.5
−0.5 1
4
1.5
2
2.5 Time (s)
3
3.5
−0.5 1
4
1.5
2
3
3.5
4
2.5 Time (s)
3
3.5
4
6cm−−LeadIII
Voltage (V)
Standard Lead System New Lead System
0.5
0
−0.5 1
4
1 Standard Lead System New Lead System
Voltage (V) 2.5 Time (s)
3.5
0
6cm−−LeadII
0
3
0.5
1
0.5
2.5 Time (s)
Standard Lead System New Lead System
0
Standard Lead System New Lead System
2
2
5cm−−LeadIII
0.5
6cm−−LeadI
1.5
1.5
1
Voltage (V)
Voltage (V)
Voltage (V)
−0.5 1
4
Standard Lead System New Lead System
1
Voltage (V)
3.5
0
1 Standard Lead System New Lead System
−0.5 1
3
0.5
5cm−−LeadII
1
−0.5 1
Standard Lead System New Lead System Voltage (V)
Voltage (V)
Voltage (V)
Standard Lead System New Lead System
−0.5 1
4cm−−LeadIII
1
1.5
2
2.5 Time (s)
3
3.5
4
0.5
0
−0.5 1
1.5
2
2.5 Time (s)
3
3.5
4
Fig. 4 The comparison of signals between standard placement and new placements
and 50 Ω Inverted F PCB antenna (IFA). It receives the data from sensor node and sends back the acknowledgment message with Tx output power adjustment suggestion. The received data is transmitted to FT220X and then sent to PC. The detailed process of the communication between sensor node and coordinator is shown in Fig. 7.
For the purpose of low power consumption of the transceiver, this work employed a dynamic power adjustment method [21]. Every data packet the coordinator receives includes the RSSI. RSSI, which ranges from 0 dBm to -100 dBm, is computed internally in the radio by averaging the signal power over eight symbol periods of the incoming packet. It is related to the link state (distance and obstacle) between sensor node and coordinator. Therefore, it is a good indicator to adjust the Tx output power of the sensor node. With the purpose to find a suitable method to adjust the Tx output power, an experiment is designed. Because distance is an important factor to affect the RSSI value, the Tx output power is recorded at every distance, as shown in Fig. 8. In order to get the same signal strength at every distance, the output power is adjusted until the RSSI reaches between -60 dBm and -70 dBm. To avoid other factors affecting RSSI, the measurement is done in an open area. The result indicates that the Tx output power does not change linearly, but changes faster in lower levels than in higher levels. The relationship between Tx output power and current consumption of transceiver is listed in Table 2. It states that it can
Dynamic transmission power control In the most of the body sensor network system, the transceiver always consumes most of the power of sensor node. In the body sensor application, the patient who is carrying the sensor will move in a certain range. To successfully perform over a long term and monitor signal continuously, the transmitter power should not be set in a fixed level, but be adjustable according to the received signal strength. At night, the patient would stay at the same place. If the coordinator is put close to the sensor, and the Tx output power of the sensor is adjusted to a lower level, the power consumption can be significantly reduced. Fig. 5 System diagram
RA INA
BPF
LA
A INA
LL
BPF
D
MCU
RX TX
RX TX
MCU
35
J Med Syst (2015) 39:35
Page 6 of 10 5
Tx Output Power (dBm)
0 −5 −10 −15 −20
Fig. 6 The design of the sensor node −25
save 11 mA in lowest level compared with the highest level. But at the lower level, the current consumption does not change a lot. According to these two results, the Tx output power is set at 9 levels (as listed in Table 3) to achieve fast adjustable performance. It adjusts fast at low levels but slowly at high levels. This setting can largely avoid the link break between sensor node and coordinator, when the link state varies quickly. The previous measurement of this work also shows that when the RSSI becomes lower than -90 dBm, the data can not be received successfully. As a result, the rules of the adjustment are set as: RSSI ≤ −80dBm : PowerLevel ¼ 8 −80dBm < RSSI < −75dBm : PowerLevel þ 1 RSSI > −65dBm : PowerLevel−1
Wake up every 1s
No
ð2Þ
0
5 10 15 20 Distance between sensor node and coordinator (m)
Fig. 8 Transmission power changes according to distance
With these rules, the RSSI is controlled between -65 dBm and -75 dBm. The adjustment suggestion will be sent with the acknowledgment to the sensor node. The detailed information of the communication is shown in Fig. 7.
GUI data acquisition host A GUI Data Acquisition Host is built to set the UART parameters and display the signals received from the Coordinator. The QRS complexes can also be detected at real time using the Hamilton-Tompkins algorithm [22], and the heart rate is calculated according to the period between QRS complexes.
Sensor Node
Coordinator
Initialization Search coordinator
Initialization Wait for connection
Find coordinator? Match network key? Yes
Enter into deep sleep mode (Power Mode 2)
Send connection ask Wait for answer No
No
Yes Connection allowed Send response
Succesful? Yes
Wait for data
Enable ADC Interruption ADC interruption, Wake up No Enter into sleep mode (Power Mode 1)
Sending task? Yes Send the data Wait for answer
No Yes
3 times?
No
Receive succesfully?
Read RSSI Send acknowledgement with Tx power adjustment suggestion
Yes
Fig. 7 Flow chart of communication between sensor node and coordinator
No
Yes
Send succesfully?
Adjust the Tx power
25
Send data to PC
J Med Syst (2015) 39:35 Table 2
Page 7 of 10 35
Current consumption of CC2530 at every Tx output power [18]
Tx output power (dBm)
4.5
2.5
1
-0.5
-1.5
-3
-4
-6
-8
-10
-12
-14
-16
-18
-20
-22
Current consumption (mA)
34
31
29
28
27
27
26
26
25
25
25
25
25
24
24
23
Implementation and measurement results
Dynamic transmission power control
ECG monitoring system
The best way to evaluate the performance of dynamic transmission power control is to do the measurement under body motion. As the distance is the main factor affecting the adjustment, the measurement is designed as follows: the volunteer, who wears the sensor node, starts walking away from the coordinator, and then turns back at 20 m and walks towards to the coordinator with different speeds. The fluctuations of Tx output power level and RSSI value during this procedure are recorded (as shown in Fig. 12). The results illustrate that, no matter the volunteer walks fast or slow, the power level can be adjusted automatically and the RSSI is kept around -70 dBm. During walking under normal speed, about 20 % power is saved in 45 s. As exhibited in Fig. 12d, when the volunteer is under rest, the power level and RSSI can be kept stable and 30 % power is saved. This strongly demonstrates that, with the dynamic transmission power control, the power consumption can be significantly reduced when the patient is sleeping during night or remaining motionless. Figure 12 also indicates that when the volunteer is walking away and towards to coordinator, the power level does not change linearly, but with some fluctuations. This phenomenon is caused by the surrounding environment. The power level is adjusted according to the RSSI. But the RSSI not only depends on the distance, but also on the surrounding environment. The walls and other obstacles will also affect the RSSI. Consequently, when there are some obstacles between the
Figure 9a shows the photograph of sensor node, the size of which is controlled in 5.5 cm×2.5 cm. Bottom view is the analog front end and the connectors to electrodes. Top view shows the rechargeable battery and the charging port. The transceiver and the antenna are implanted below the battery. Figure 9b shows the photograph of the coordinator. The maximum communication distance between sensor node and coordinator is 30 m, which can easily fulfill home usage applications. The current is measured when there is no connection with the coordinator. As described before, the sensor node enters into deep sleep mode during this situation and consumes about 5 mA current. After every 1 s it wakes up and searches the coordinator. The current consumption of this process is clearly shown in Fig. 10a. As shown in Fig. 10b, the current consumption of sensor node under communication with coordinator is measured. It can be apparently seen that, the data is sent every 100 ms. During every period, three current peaks, which denote sending data, receiving acknowledgment and sending successful feedback respectively, are achieved. The period between the ripples tangibly reveals that, the CC2530 enter into sleep mode during the task free time. Owing to the power mode control, the sensor node saves plenty of power, especially when there is no connection with the coordinator. Even if the sensor works continuously, the battery life can be more than 52 h. The GUI data acquisition host is shown in Fig. 11. It can successfully detect the QRS complexes and calculate the heart rate at real time. The sensitivity of QRS complex detection is 97.22 % when the body is peaceful and 91.25 % when the body is running. The GUI can also show an alarm when the heart rate is abnormal. The whole system with compact size and light weight establishes stable communication among sensor node, coordinator and GUI. Along with that, high quality signals are achieved even under body motion (walking and running). Table 3
Tx output power level settings
Power level
0
1
2
3
4
5
6 7
8
Tx output power (dBm) -22 -18 -8 -4 -1.5 -0.5 1 2.5 4.5
(a) Sensor node
(b) Coordinator
Fig. 9 Photograph of the system. (a) Sensor node (b) Coordinator
35
J Med Syst (2015) 39:35
Page 8 of 10 40
Current (mA)
30
20
10
0
1
2
3
4
5
6
Time (s)
(a) During coordinator absent 40
The fluctuations in low levels are stronger than in high levels, because the adjustment steps in low levels are bigger than in high level. It can be avoided by reducing the steps. According to the measurement in Fig. 12, the Tx output power varies very fast in lower distances. In consequence, small steps in low levels will cause the link disconnection when the link state varies quickly, especially when the sensor moves very fast away from the coordinator. If the Tx output power cannot be adjusted fast enough, the data cannot be received successfully. To keep the stability of the link, big steps in low power levels are still used in adjustment.
Current (mA)
30
20
Discussion and conclusion
10
Discussion 0 50
100
150
200
250 300 Time (ms)
350
400
450
(a) Novel limb electrode placements
(b) During communication with coordinator Fig. 10 Current consumption. (a) During coordinator absent (b) During communication with coordinator
sensor and coordinator, the power level will also increase, even though the distance is very short.
Fig. 11 GUI data acquisition host
Compare to the design in [23–26], the novel placements for limb electrodes presented in this paper is possible to connect the electrodes with sensor node completely wireless. This system has more flexibility and compactness for the patients than the design in [3, 27]. It only needs one sensor node in small size and avoids the time synchronization between every
J Med Syst (2015) 39:35
Page 9 of 10 35
−60
2
−70
−80
0
−80
−90 30
−2
−60
2
−70
0
20
25
0
5
10
15
(a) Fast walking
Tx Output Power Level
−60
2
−70
−80
0
−80
−90 60
−2
−60
2
−70
0
50
−40
4
4
40
−30
−50
−50
30 Time (s)
−90 45
6
6
20
40
Tx Power Level RSSI
8
−40
RSSI (dBm)
Tx Output Power Level
8
10
35
10
−30 Tx Power Level RSSI
0
30
(b) Normal walking
10
−2
20 25 Time (s)
0
5
(c) Slow walking
10
15
20 Time (s)
25
30
35
RSSI (dBm)
15 Time (s)
RSSI (dBm)
4
4
10
Tx Output Power Level
Tx Output Power Level
−50
−50
5
−40
6
6
0
−30 Tx Power Level RSSI
8
−40
RSSI (dBm)
Tx Power Level RSSI
8
−2
10
−30
10
−90 40
(d) Resting
Fig. 12 Power level adjustment and RSSI during body motion and resting. (a) Fast walking (b) Normal walking (c) Slow walking (d) Resting
sensor node. Although it is bigger in size than [9, 10, 13], the proposed system can detect more lead signals and supply more information for diagnosis.
defined. It can be adjusted much faster than the 32 levels in [21], and avoid the disconnection when the link state varies rapidly.
(b) Wireless ECG monitoring system Based on the new electrode placements, the size of the sensor is controlled in 5.5 cm×2.5 cm, which can satisfy the long term wearable requirement. Compared to the design in [26–28], this system uses simplified circuit, but can still get the high quality signals and meanwhile make the size of the sensor compact. Unlike the design in [13], this design can monitor the ECG signals continuously. Without the hand held device (HHD) which is designed in [29], the system becomes much more flexible. (c) Dynamic transmission power control Many transmission power control protocols have been employed in wireless sensor networks [30]. However most of these studies have been targeted for static networks [31, 32]. The studies [14] and [21] published their researches on transmission power control in body area network. Compare to [14], the proposed work simplified the adjustment protocol according to our practical application. Based on the measurement of the Tx output power at every distance and the current consumption at every Tx output power, 9 power levels are
Conclusion This paper introduces a wireless and wearable ECG monitoring system for long term homecare. The novel electrode placements account for the possiblity of compact size and long term wearable sensor design. It can be successfully applied in the wireless and wearable ECG sensor system. The sensor node is implemented in small size and with power control. The communication between the sensor node and coordinator performs securely and steadily. The power mode control on sensor node substantially reduces the power consumption during the task absence time. It tremendously extends the life time of battery. The dynamic transmission power control is employed. The Tx output power of the transceiver on sensor node is automatically adjusted according to the RSSI value. This method is tested under both the fast and slow body motion. The results demonstrate that, the adjustment works agilely, accurately and stably. It reduces 20 % power consumption during normal walking and 30 % during resting. Obviously, the dynamic transmission power control will contribute a lot in extending the battery life time. Overall, the whole system can certainly
35
Page 10 of 10
satisfy the long term homecare usage and provide more physical mobility to patients.
References 1. Wang, Y., Yu, K., Wang, D., Zhao, C., Wang, L. and Wang, P., BMulti-Model Diagnosis Method for Lung Cancer based on MOSSAW Breath Detecting e-Nose,^ in Proceedings of The 14th International Symposium on Olfaction and Electronic Nose, 2011. 2. Wang, Y., Hu, Y., Wang, D., Yu, K., Wang, L., Zou, Y., Zhao, C., Zhang, X., Wang, P., and Ying, K., The analysis of volatile organic compounds biomarkers for lung cancer in exhaled breath, tissues and cell lines. Cancer. Biomark. 11:129–137, 2012. 3. Cao, H., Li, H., Stocco, L., and Leung, V. C. M., Wireless three-pad ECG system: challenges, design, and evaluations. Commun. Net. J. 13:113–124, 2011. 4. Einthoven, W., Weiteres ueber das elektrokardiogram. Pflueger Arch. ges. Physiol. 122:517–584, 1908. 5. Wilson, F. N., Johnston, F. D., Rosenbaum, F. F., Erlanger, H., Kossmann, C. E., Hecht, H., Cotrim, N., Oliveira, R. M. d., Scarsi, R. and Barker, P. S., BThe precordial electrocardiogram,^ AHJ, vol. 27, pp. 19-85, 1944. 6. Mason, R. E., and Likar, I., A new system of multiple-lead exercise electrocardiography. AHJ. 71:196–205, 1966. 7. Dower, G. E., Yakush, A., Nazzal, S. B., Jutzy, R. V., and Ruiz, C. E., Deriving the 12-lead electrocardiogram from four (EASI) electrodes. J. Electrocardiol. 21:182–187, 1988. 8. Munshi, M., Xu, X., Zou, X., Soetiono, E., Teo, C. S. and Lian, Y., BWireless ECG plaster for body sensor network,^ in Proceedings of ISSS-MDBS 2008, 2008. 9. Masse, F., Penders, J., Serteyn, A., Bussel, M. van and Arends, J., BMiniaturized wireless ECG-monitor for real-time detection of epileptic seizures,^ in Proceedings of Wireless Health 2010, 2010. 10. Altini, M., Polito, S., Penders, J., Kim, H., Helleputte, N. Van, Kim, S. and Yazicioglu, F., BAn ECG patch combining a customized ultralow-power ECG SoC with bluetooth low energy for long term ambulatory monitoring,^ in Proceedings of the 2nd Conference on Wireless Health, 2011. 11. Hoekema, R., Uijen, G., and van Oosterom, A., On selecting a body surface mapping procedure. J. Electrocardiol. 32:93–101, 1999. 12. Torfs, Yazicioglu, T. R., Kim, S., Kim, H., Hoof, C. Van, Buxi, D., Romero, I. a. W. J., Masse, F. and Penders, J., BUltra low power wireless ECG system with beat detection and real time impedance measurement,^ in Proceedings of IEEE BioCAS 2010 Conference, 2010. 13. Gomez-Clapers, J., and Casanella, R., A fast and easy-to-use ECG acquisition and heart rate monitoring system using a wireless steering wheel. Sens. J. IEEE. 12:610–616, 2012. 14. Kim, S., and Eom, D., Link-state-estimation-based transmission power control in wireless body area networks. IEEE J-BHI. 18: 1294–1302, 2014. 15. Wang, Y., Wunderlich, R. and Heinen, S., BDesign and Evaluation of a Novel Wireless Reconstructed 3-Lead ECG Monitoring System,^ in Proceedings of IEEE BioCAS 2013 Conference, 2013. 16. Ambu® BlueSensor M. Ambu Inc., 2013.
J Med Syst (2015) 39:35 17. Wang. Y., Wunderlich. R. and Heinen. S., BA low noise wearable wireless ECG system with body motion cancellation for long term homecare,^ in Proceedings of IEEE Healthcom 2013 Conference, 2013. 18. A true system-on-chip solution for 2.4-GHz IEEE 802.15.4 and ZigBee applications. Texas Instruments Inc., 2011. 19. Andersen, A., 2.4 GHz inverted F antenna. Texas Instruments Inc., 2008. 20. USB 4-BIT SPI/FT1248 IC FT220X. Future Technology Devices International Ltd., 2014. 21. Xiao, S., Dhamdhere, A., Sivaraman, V., and Burdett, A., Transmission power control in body area sensor networks for healthcare monitoring. IEEE J-SAC. 27:37–48, 2009. 22. Pan, J., and Tompkins, W. J., A real-time QRS detection algorithm. IEEE Trans. Biomed. Eng. BME-32:230–236, 1985. 23. Burns, A., Doheny, E., Greene, B., Foran, T., Leahy, D., O’Donovan, K. and McGrath, M., BSHIMMER™: an extensible platform for physiological signal capture,^ in Proceedings of Annual International Conference of IEEE EMBC 2010, 2010. 24. Gaxiola-Sosa, J., Mohsin, N., Palliyali, A., Tafreshi, R. and Entesari, K., BA portable 12-lead ECG wireless medical system for continuous cardiac-activity monitoring,^ in Proceedings of MECBME 2014, 2014. 25. Gao, H., Duan, X., Guo, X., Huang, A. and Jiao, B., BDesign and tests of a smartphones-based multi-lead ECG monitoring system,^ in Proceedings of Annual International Conference of IEEE EMBC 2013, 2013. 26. Tan, T.-H., Chang, C.-S., Huang, Y.-F., Chen, Y.-F., and Lee, C., Development of a portable linux-based ECG measurement and monitoring system. J. Med. Syst. 35:559–569, 2011. 27. Winokur, E., Delano, M., and Sodini, C., A wearable cardiac monitor for long-term data acquisition and analysis. IEEE Trans. Biomed. Eng. 60:189–192, 2013. 28. Chen, H., Wu, W., and Lee, J., A WBAN-based real-time electroencephalogram monitoring system: design and implementation. J. Med. Syst. 34:303–311, 2010. 29. Fensli, R., Dale, J., O’Reilly, P., O’Donoghue, J., Sammon, D., and Gundersen, T., Towards improved healthcare performance: examining technological possibilities and patient satisfaction with wireless body area networks. J. Med. Syst. 34:767–775, 2010. 30. Pantazis, N., and Vergados, D., A survey on power control issues in wireless sensor networks. IEEE Commun. Surv. Tutor. 9:86–107, 2007. 31. Zhou, G., He, T., Krishnamurthy, S. and Stankovic, J. A., BImpact of radio irregularity on wireless sensor networks,^ in Proceedings of MobiSys 2004, 2004. 32. Zhao, J. and Govindan, R., BUnderstanding packet delivery performance in dense wireless sensor networks,^ in Proceedings of the 1st International Conference on Embedded Networked Sensor Systems, 2003. 33. Malmivuo, J., and Plonsey, R., Bioelectromagnetism: principles and applications of bioelectric and biomagnetic fields. Oxford University Press, New York, 1995. 34. Yanowitz, F. G., Introduction to ECG interpretation V8.0. Eccles Health Sciences Library University of Utah, Salt Lake City, 2012. 35. Welinder, A., Sörnmo, L., Field, D. Q., Feldman, C. L., Pettersson, J., Wagner, G. S., and Pahlm, O., Comparison of signal quality between easy and Mason-Likar 12-Lead electrocardiograms during physical activity. AJCC. 13:228–234, 2004.
