Journal of Medical Systems

, Volume 34, Issue 3, pp 303–311

A WBAN-based Real-time Electroencephalogram Monitoring System: Design and Implementation

Authors

  • Haifeng Chen
    • Graduate School of Computer Science and EngineeringPusan National University
  • Wanqing Wu
    • Graduate School of Computer Science and EngineeringPusan National University
    • Graduate School of Computer Science and EngineeringPusan National University
Original Paper

DOI: 10.1007/s10916-008-9242-9

Cite this article as:
Chen, H., Wu, W. & Lee, J. J Med Syst (2010) 34: 303. doi:10.1007/s10916-008-9242-9

Abstract

In this study, a flexible wireless body area network (WBAN) node platform has been designed and implemented based on the Zigbee technology. In order to provide wide range WBAN for health monitoring, a Zigbee/Internet Gateway (ZiGW) has also been developed rather than using a PDA or a host PC to connect different WBANs by using the Internet as the communication infrastructure. The proposed body sensor node platform promises a cost-effective, flexible platform for developing physical sensor node in real-time health monitoring. The ZiGW can provide an effective method to connect WBAN with the Internet. In this work, we present the implementation of an Electroencephalogram (EEG) monitoring system using the proposed methods. In this proposed system, real-time EEG signals can be remotely monitored by physicians via Internet, and the collected EEG data is stored in the online EEG database which can be shared with physicians or researchers for further analysis.

Keywords

Wireless body area networkSensor node platformZigbee/Internet gatewayEEG conditioningEEG monitoring

Introduction

Advances in wireless communication, sensor design, and energy storage technologies make Wireless Sensor Network (WSN) with pervasive concept rapidly becoming a reality [1]. Pervasive health or patient monitoring systems integrated into a telemedicine system are novel information technology that will be able to support early detection of abnormal conditions and prevention of serious consequences [2].

Recently, WSN is becoming a promising technology for various applications. One of its potential deployments is the Wireless Body Area Network (WBAN) for measuring physiological signals to implement health monitoring. In recent years, many WSN node platforms have been developed, such as Berkely’s Mica [3] and Moteiv’s Telos [4], ETH’s BTnodes [5], and Intel’s iMote [6]. A BSN node prototype is developed in Ubimon project [7]. Unfortunately, most of these platforms are designed for network research, or environment monitoring. As the features of the physiological signals monitoring are different from environment monitoring, a WBAN specified sensor node platform needs to be developed in order to facilitate the research of WBAN. Miniature sensor nodes monitoring multi-channel physiological signals will be one of the most challenging tasks for designers. The physiological signal from a human body is usually weak and coupled with inner or outer noise. Thus, the signal should go through an amplification process to increase its strength. It then passes through a band pass filter to remove noises and through an Analog/Digital Converter (ADC) stage to be converted into digital signal for further digital processing. Moreover, sensor nodes must be energy efficient in order to avoid frequent battery recharge as they are battery powered.

In addition, some Internet based patient monitoring systems have been reported in recent years in [810], and some wireless sensor networks based on personal health monitoring systems have also been introduced in [1113]. With the advancement of wireless sensor networks (WSN), some WBAN based systems for health monitoring are also introduced by many researchers recently in [1417]. On the other hand, the wireless EEG device solutions are also introduced in [1820].

However, most of the current efforts have mainly focused on the devices that are using the commercial off-the-shelf sensor node platform to develop WBAN sensor nodes. This will incur a high cost for the WBAN sensor node with low integration. Additionally, most of these systems are monitoring sensor signals only from a single patient’s body. Monitoring many physiological signals from a large number of patients in real time is one of the most important requirements in order to deploy a complete WBAN based network system in medical centers. Such an application shows some challenges in both hardware and software designs. This work makes full use of the Internet to deploy a WBAN based physiological signal monitoring system in real time via Internet. We also propose a new WBAN node platform and a solution which can support the WBAN to seamlessly connect to the Internet in this paper.

In this paper, we introduce a new sensor node platform for WBAN, and a Zigbee/Internet Gateway (ZiGW) which is capable of supporting wide ranges of online medical monitoring applications via Internet. Moreover, an online database server is designed to save acquired data and users’ information for further analysis, and a client program has been designed for physicians or researchers to do online monitoring and data analysis. The architecture of proposed Internet connectible WBAN is illustrated in Fig. 1. The tiny common WBAN node platform can be easily integrated with many kinds of biosensor through its provided sensor connector. Physiological signals, such as EEG, ECG, SpO2, blood pressure, etc., measured by these sensor nodes are gathered by body area network coordinator wirelessly. The coordinator can be a portable wireless device (e.g. PDA) or a WBAN node itself. Then, the coordinator transfers the gathered data to an online database via the proposed ZiGW. The measured physiological data can be shared with physicians for further analysis. In this work, the coordinator is the zEEG itself which integrated with an 8-channel EEG sensor for a preliminary evaluation. This work is also based on our former portable EEG device development projects which are described in [2123].
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig1_HTML.gif
Fig. 1

Architecture of the proposed WBAN based physiological signal monitoring system

The rest of the paper is organized as follows. “Materials and methods” proposes the design of our new sensor node platform and ZiGW for wide range WBAN and the design of EEG sensor is also introduced. “Results” describes the proposed system implementation and experiment results in details. Finally “Conclusion” concludes this paper.

Materials and methods

Design of WBAN node platform and zEEG

The WBAN, unlike wired monitoring systems, provides long term and continuous monitoring of patients under their natural physiological states even when they move. Therefore, reliable EEG signals can be measured with the EEG sensor node in the proposed WBAN based physiological signal monitoring system. Figure 2 shows the architecture of zEEG, which consists of 8-channal EEG sensor with a common WBAN node platform. In this design, the EEG sensor, which acts as the EEG signal conditioning part, can be inserted into the developed WBAN node platform using predefined common connector. The main parts of it are listed and introduced as follows.
  1. a)

    EEG sensor: As mentioned above, physiological signals are usually week and easy to be interfered by undesired noises (EEG signals typically have an amplitude in the range of 10 µ ~ 100 µV). Therefore, both amplifying and filtering are required for further signal processing. Moreover, human skin typically provides source impedance on the order of 1 M ~ 5 Mohm. So amplifiers must match the source impedance or have greater input impedance than the source skin impedance to acquire surface physiological signals. To avoid 60 Hz (or 50 Hz) power line noise, a narrow band notch filter is also needed for signal processing. For these reasons, a relatively high common mode rejection ratio (CMRR) is necessary to reject common mode signals. Thus, a high input impedance, high CMRR and moderately high gain instrumentation amplifier will be a good choice as the differential amplifier of the EEG conditioning circuit. In this work, the instrumentation amplifier chip named INA128AIM [24], is selected, which is a low power and general purpose instrumentation amplifier with low input bias current. It also features a high CMRR of 120 dB and a differential input impedance of 10 Gohm║2 pF which can satisfy the required conditions well.

     
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig2_HTML.gif
Fig. 2

Block diagram of Zigbee-based EEG sensor node

In order to acquire the most useful EEG band, the cutoff frequencies of the Low Pass Filter (LPF) and the High Pass Filter (HPF) are set to 1.5 Hz and 50 Hz, respectively. The output signal of instrumentation amplifier is transferred to the 4th-order Bessel LPF with 0 dB gains. After LPF, the signal is then filtered by the 2nd-order Butterworth HPF to attenuate low frequency signal with 40 dB gains. Then, the signal goes into the 60 Hz notch filter to reject the 60 Hz power line noise. Finally, the signal is amplified by the 2nd invert amplifier with 12 dB gains. The operation amplifier (OpAmp) chip used here is LMC6464AIM from National Semiconductor [25], which is a low power operation amplifier with Rail-to-Rail input and output. The structure of the signal conditioning circuit is shown in Fig. 3., and the specifications of zEEG’s signal conditioning part are given in Table 1.
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig3_HTML.gif
Fig. 3

Structure of EEG conditioning circuit

Table 1

Specifications of EEG conditioning part

Part name

Character

Gain

Pre-amplifier

Differential input impendence: 10 Gohm║2 pf

25 dB

CMRR: 120 dB

input bias current: 1 nA

input offset voltage: 25 µV

LPF

4th-order Bessel filter

0 dB

cutoff frequency: 50 Hz

HPF & amplifier

2nd-order Butterworth filter

40 dB

cutoff frequency: 1.5 Hz

Invert amplifier

Notch filter

60 Hz Twin-T notch filter

0 dB

2nd amplifier

Invert amplifier

12 dB

  1. b)

    Zigbee transceiver: The Zigbee transceiver of zEEG features the Chipcon CC2420 chip for wireless communications [26]. The CC2420 is an IEEE 802.15.4 compliant radio providing the PHY and some MAC functions. The CC2420 is highly configurable for many applications with the default radio settings providing IEEE 802.15.4 compliance. The CC2420 is controlled by the microcontroller through a 4-wire SPI bus and a series of digital I/O lines as shown in Fig. 4. FIFO and FIFOP status pins can be used to interface to the receiving and transmitting FIFOs, the CCA pin is for clear channel assessment, and the SFD pin can be used for timing information. The CC2420 has programmable output power and provides a digital received signal strength indicator (RSSI) that transferred with data frame together.

     
  2. c)

    Microcontroller: The core of our proposed WBAN node platform is an 8051 instruction compatible microcontroller featuring 2 kB on-chip RAM, 64 kB of flash ROM, and 2 kB of EEPROM for information storage. The 8051 instruction compatible microcontroller is widely used by industries that will be familiar to developers.

     
  3. d)

    Antenna: The Antenova 50 ohm Mica chip antenna is selected to ensure high performance, minimal power consumption and small dimensions for the node platform [27]. The Antenova Mica chip antenna is easy to integrate, and is engineered specifically for Zigbee devices operating at 2.47 GHz with an average efficiency of 65%.

     
  4. e)

    Power supply and management: zEEG is powered by a 9 V battery temporarily. And the battery status is monitored by using the CC2420 provided on-chip battery monitoring function. We are now considering a small sized Li-Ion rechargeable to replace the 9 V battery.

     
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig4_HTML.gif
Fig. 4

Microcontroller interface of CC2420

Design of Zigbee/Internet Gateway (ZiGW)

This study aims at developing a WBAN sensor node platform for physiological signals sensing to implement a real-time EEG monitoring system. As the Internet is a wide spread and inexpensive communication infrastructure, the Internet would be a good choice to connect the scattered WBANs together. Then, a gateway between WBAN and Internet is required. The architecture of our proposed ZiGW is shown in Fig. 5.
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig5_HTML.gif
Fig. 5

Block diagram of the proposed Zigbee/Internet gateway

Zigbee transceiver of ZiGW also features ChipCon CC2420 chip for communication with sensor node wirelessly. Here, received Zigbee data frames are needed to encapsulate with TCP/IP header. To provide a high TCP/IP processing speed, a hardwired TCP/IP stack chip, named W3100A from WIZnet [28] which is performed in our previous study, is used to do the TCP/IP packet processing. Figure 6 shows the protocol stack that integrated in this chip, such as Internet Protocol (IP), Address Resolution Protocol (ARP), Internet Control Messages Protocol (ICMP), Transmission Control Protocol (TCP), and Media Access Control (MAC), etc. With this chip, system integration is convenient and OS is not needed. As the Ethernet PHY is not integrated in W3100A, an external Ethernet PHY chip RTL82018 from REALTEK is used in ZiGW and it connects the W3100A through the Media Independent Interface (MII). After the received Zigbee frame is encapsulated with TCP/IP header, the packet will be transferred to database server directly via Internet.
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig6_HTML.gif
Fig. 6

Block diagram of hardwired TCP/IP chip W3100A

Design of database and software

In this sub-section, software system of the proposed WBAN will be introduced. The system consists of three main parts, the EEG database, the management program, and the client viewer. As a similar system was also described in our former work in [23], here we will introduce these parts in brief.
  1. a)

    EEG database: It is an online EEG database storing the monitored EEG data from patients, and it is shared with physicians or researchers to monitor or analyze EEG data. Authorized users can access it in anywhere at any time via Internet by using our proposed client Viewer program. Moreover, the information of devices (zEEG and ZiGW) and users are also stored in the implemented database for security purposes.

     
  2. b)

    The Management program: Because there are a number of zEEG nodes connected with server via several ZiGWs, a program is needed to manage these nodes for administrator in the server side. This program also provides clients’ management. With this program, the administrator can view the status of the connected zEEG nodes and Viewer Client via online.

     
  3. c)

    Viewer Client: This is a client program designed for physicians or researchers to access the online database and analyze retrieved EEG data. Some basic signal processing methods are provided in this program, such as power spectrum analysis, digital filters, and so on.

     

Results

zEEG sensor node platform

The first version of the proposed sensor node platform for WBAN has been implemented, as shown in Fig. 7. EEG sensor node (8 channels) has a size of 55 × 607 mm, and WBAN node platform board has a size of 65 × 80 mm. The zEEG case dimension is 70 × 130 × 23 mm include batter. It is not specified for EEG sensors. It is also ideal to serve as a platform for other biosensors, such as ECG sensor, EMG sensor and so on. The proposed WBAN node platform can significantly cut the development cycle and the cost for wireless biosensor node development. In this paper, we only discuss the zEEG (WBAN node platform with EEG sensor) implementation and experiment results.
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig7_HTML.jpg
Fig. 7

Implemented zEEG in case (left one is EEG sensor integrated with WBAN node platform; right-top one is EEG sensor, and right-below shows the WBAN node platform with connected EEG electrodes)

An analysis of the results for the EEG conditioning part will be introduced firstly as it is important to acquire clean EEG signals. Figure 8 shows the Amplitude Frequency Response (AFR) of the total eight channels of zEEG. The AFR is measured manually from 0.4 Hz to 100 Hz by using an oscilloscope for frequency measurements and a function generator was used as a signal input. In Fig. 8, the pass band of EEG conditioning circuit is from 1.5 Hz to 50 Hz at −3 dB which is consistent with our design goals. About −20 dB attenuation rate is achieved at 60 Hz and the gain error of each channel is less than 0.05 in the pass band of zEEG.
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig8_HTML.gif
Fig. 8

Amplitude Frequency Response (AFR) of EEG conditioning part

In addition, we also adopt another method to evaluate the performance of the zEEG. A commercial EEG device named QEEG-4 from LAXTHA Inc. [29] is used to compare the output EEG data with our zEEG under the same conditions. In this experiment, QEEG-4 and zEEG share the same Ground, Reference and measure electrodes as illustrated in Fig. 9. Sampling rate of the two devices is set to 256 Hz, and EEG signals of two mental conditions (close eyes and open eyes) were acquired in this experiment. The specification of QEEG-4 and zEEG is introduced in Table 2.
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig9_HTML.gif
Fig. 9

Experiment environment setup for comparison with QEEG-4

Table 2

Specifications of QEEG-4 and zEEG

Device

QEEG-4

zEEG

Operational input voltage

±5 V

±5 V

Lower frequency response

1.5 Hz

1.5 Hz

Upper frequency response

45 Hz

50 Hz

Output signal range

<5 V

±2.047 V

ADC resolution

8 bits

12 bits

Figure 10 shows a segment of the acquired EEG signals from one channel of QEEG-4 and zEEG. We can find that the collected EEG waveform from the zEEG is similar with QEEG-4’s output intuitively from Fig. 9., and from 1–2 s, the Alpha wave (8 ~ 12 Hz, appearing when human is awareness and eyes are closed) is clearly appeared. Furthermore, the Fourier Transform result in Fig. 11 gives us a frequency domain view of the captured EEG signal from the two devices respectively. Figure 11 also shows zEEG achieves similar results in comparison with QEEG-4, and the power concentrates on the band of 8 ~ 12 Hz.
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig10_HTML.gif
Fig. 10

Captured EEG signals from QEEG-4 and zEEG

https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig11_HTML.gif
Fig. 11

Fourier transform result of captured EEG signals (above one is from zEEG and the below one is from QEEG-4)

Moreover, the percentage of main EEG components, such as Delta (below 4 Hz), Theta (4–7 Hz), Alpha (8–12 Hz) and Beta (13–30 Hz) wave, is widely used in EEG signal processing to analyze the state of consciousness. So, we also evaluated the percentage of main EEG components of acquired EEG signals from QEEG-4 and zEEG. The position of EEG electrode lies in the front lobe of the brain (e.g., F3 in the 10–20 International System). Four healthy volunteers were involved in this experiment and the measurement time was 3 min. The percentage of main EEG components of acquired EEG signals from QEE-4 and zEEG is shown in Fig. 12. Analysis result shows that Alpha band occupies 50% spectrum power of the whole EEG band of the two devices while subject was in eyes closed condition, and only 0.3% difference exists in the percentage of main EEG components in power spectrum between QEEG-4 and zEEG. From the comparisons, we can conclude that there is no conspicuous difference in the output analog signal from QEEG-4 and zEEG.
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig12_HTML.gif
Fig. 12

Main EEG components ratio comparison result of QEEG-4 and zEEG

Finally, low power consumption is the most challenging part for wireless sensor network. In this study, we measure the battery life during actual EEG monitoring despite the power efficient mode is not used in this preliminary evaluation. A 9 V battery with capacity of 350 mAh is used for power supply, and zEEG is in 100% active duty cycle in this experiment. Since the sampling rate is 256 Hz and one sample length is 2 bytes, the data rate for transmission of total eight channels is 32 kbps (256 × 2 × 8 × 8) without Zigbee frame header. zEEG can work about 7 h in real-time EEG monitoring under the above conditions.

Zigbee/Internet Gateway (ZiGW)

ZiGW provides the connection between WBAN and Internet as introduced in “Materials and methods”, the photography of finished PCB board is shown in Fig. 13. It is powered by a 5 V DC adapter and an omni-directional antenna which can achieve strong wireless signal sensitivity as the antenna is not a mobile unit in the proposed system.
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig13_HTML.jpg
Fig. 13

Hardware design of Zigbee/Internet Gateway (ZiGW)

In order to evaluate the performance of ZiGW, the Receive Signal Strength Indicator (RSSI) and Packet Reception Ratio (PRR) of the gateway have been measured and calculated. This experiment was deployed in indoor environment and the above two parameters were measured up to 35 m range. The PRR was recorded and calculated at each place after 1,000 packets transmission. The RSSI was the average value of the transmitted 1,000 packets. Figure 14 illustrates the average RSSI of ZiGW in each different distance position up to 35 m. Figure 15 shows the average PRR of ZiGW in different RSSI.
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig14_HTML.gif
Fig. 14

RSSI value of ZiGW in each different communication distance

https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig15_HTML.gif
Fig. 15

Packet reception ratio of ZiGW under different RSSI level

Figure 14 shows that the average value of RSSI is greater than −70 dBm when the communication distance is less than 11 m, and in Fig. 15, the measurement result shows that the PRR is greater than 99.5% while the value of the RSSI is about −70d Bm. Thus, we can conclude that the PRR of ZiGW is greater than 99.5% while the communication distance between sensor node and gateway is no more than 11 m. So, the implemented ZiGW can provide reliable short distance wireless communication for the proposed physiological signal monitoring system. On the other side, the hardwired TCP/IP chip is able to provide 20 Mbps reliable TCP data transmission, and to support the gateway with a total bandwidth of 32 kbps.

Database, software programs and monitoring

Since the distributed WBANs are communicating with the server via Internet, the information of sensor nodes and gateway of each WBAN are all predefined and stored in a database. The PAN ID and node ID, which are included in Zigbee header, are used as an identifier to identify the source of data.

In order to manage these sensor nodes, a management tool is developed for system administrator. The user interface of this tool is shown in Fig. 16a. In this program, both the connected sensor nodes and connected remote viewer users are all listed; as a result, the status of the current system can be handled by the administrator and the information of online sensor nodes can be transferred to online researchers or physicians in real time. They can select their concerned monitoring objects from the received online sensor node list. The information of connected sensor nodes are listed in the left side of Fig. 16a, as well as the connected viewers (who are researchers or physicians involved in this system) are listed in the right list box of Fig. 16a.
https://static-content.springer.com/image/art%3A10.1007%2Fs10916-008-9242-9/MediaObjects/10916_2008_9242_Fig16_HTML.gif
Fig. 16

a GUI of administrator program (left one) b Interface of client analysis program (right one)

Figure 16b shows the user interface of monitoring program, which is an online monitoring and analyzing program used by viewers. With this program, the physicians can monitor the healthy status of their patients at anywhere in anytime via Internet. Moreover, this program can access the online database to view the past records. EEG signals of four channels are just shown in this figure.

Conclusion

The proposed WBAN node platform and Zigbee/Internet gateway serve as a research platform for study and evaluation of WBANs for health monitoring. The main features of the realized monitoring system include:
  • Portable and multi-channel supported

  • Internet seamless connection

  • A flexible low cost common research platform

  • Real time monitoring and analyzing via Internet

Currently, we are developing some other biomedical sensors to monitor other useful physiological signal such as ECG and SpO2, and efficient communication protocols among WBAN node, ZiGW and server. Moreover, the next version of our WBAN node and ZiGW is under development in order to provide a more power efficient and higher performance platform for real time health monitoring.

Acknowledgement

This work was supported by the Korea Research Foundation Grant funded by Korea Government (MOEHRD, Basic Research Promotion Fund, KRF-2007-521-D00602).

Copyright information

© Springer Science+Business Media, LLC 2009