2010 3rd International Conference on Biomedical Engineering and Informatics (BMEI 2010)
Remote Healthcare Monitor System
Xiaobin Shen, Jianping Zeng, Tundong Liu
nt of Automation Xiamen University Xiamen, China
Abstract — Remote healthcare has the potential to greatly impact many aspect of medical care. By outfitting patients with wireless, wearable vital sign sensors, collecting detailed real-time data on physiological status can be greatly simplified. Medical care has its special needs. It needs node mobility, a wide range of data rates and high degrees of reliability, and security. This paper describes our experience with developing a combined hardware, software, and communication protocol for such a system. In Particular, this paper discusses the design and implementation of such a system in details. Keywords-wireless sensor network; WSN; remote healthcare; monitor; telemonitoring
Shenzhen Institutes of Advanced Technology Chinese Academy of Sciences Shenzhen, China Community doctors implement the treatment plans. Doing so, treatment efficiency is greatly enhanced. Current telemonitoring researches focus on either the telemonitoring solution or the data transferring method, there is only a few researches addressing the management problem of the heterogeneous systems and the low power need. The TinyOs project started at Berkeley adopted a component-based architecture, the code for new functions can be easily integrated into the source tree of the operating system, and a special designed compiler is used for generating efficient binary code. It targets at wireless sensor network with severe memory constraints, dynamic resource allocation is not supported by the system. PC applications such as labview and simulink provide the ability to assemble different processing component visually. They are designed for a user interactive environment rather than a remote system running in an unsupervised mode, remote configuration and testing of a running system at the component level is not supported. In this paper, we will present a system with low power design in mind and a model for signal processing procedure in telemonitoring device, which allow different signal processing component being assembled into a running system. The paper is organized as follows. System architecture is described in section II, and software architecture is proposed in section III, Section VI is the discussion. II. SYSTEM ARCHITECTURE
With the population aging, there is a gradual increase in age-related diseases such as heart failure, coronary heart disease, stroke, pulmonary heart disease, high blood pressure. These diseases have the following characteristic: (1) Chronic diseases. A variety of triggers can cause acute-onset, and during the onset, patients often require a long time dynamic observation. (2) Patients may be difficult or unwilling to go to hospital for treatment because of physical inconvenience or psychological reasons. These often result in delays in diagnosis and treatment and have a negative impact on the survival rate. (3) Some patients may have a variety of complications and risk factors. In the early onset of diseases, they need dynamic monitoring physiological data to help define the direction of diagnosis and treatment. (4) A considerable number of patients have a relative long history. Their doctor is familiar with them. Because of this, remote healthcare monitoring is a good option for treatment adjustments. (5) Patients with acute onset may be affected by climate, environment and the impact of the seasonal fluctuations. These often increase tension on hospital beds in a particular time. Therefore, if we can take care of patients whose are relatively stable or action inconvenience at home, we can not only save patients’ medical costs, improve the quality of treatment and life, reduce diagnostic time, but also can ease the pressure on hospital beds and improve medical services covered by the number of patients. On the other hand, with the promotion of patient-centered health community, the distance between patients and medical services are shorten, but some of the key process for the diagnosis and treatment program need a higher level involvement of large hospital doctors. By using remote healthcare monitoring, great hospital doctors and community doctors can collaborate in the process of the treatment. Great hospital doctors determine the treatment plan based on patients’ medical history and monitoring data.
According to the existing features and development trend, we propose a three-tier architecture, which is “Monitor – Home gateway – Central station”, as show in Figure 1. In this architecture, home gateway is the core of this system.
Internet Wireless sensor
Monitor Monitor Figure 1. System Architecture
Various monitors worm on body firstly collect vital signs, such as ECG, respiration, blood oxygen, blood pressure and so on. Then the collected data is transmitted via wireless sensor network to home gateways. Home gateway processes these
978-1-4244-6498-2/10/$26.00 ?2010 IEEE
data and then sends them to central station through Internet without distortion. Central station storage these data and visualize the data for further medical analysis and diagnosis. Intelligent central station will also have aided diagnosis, intelligent diagnostic functions. III. WIRELESS SENSOR NETWORK
new functionalities and modules, or if modifying existing ones. ? Interoperability – Modules must be not have any blocking functions. This makes it possible to allow other modules to be kept up-to-date.
Software Architecture is shown in Figure 4.
The system’s wireless sensor network is mainly composed by terminal monitor, wireless communication module and home gateway. Star network structure is used, as show in Figure 2, where TX is the transmitter, and RX is the receiver. This structure is used for centralized control. Failure on enduser equipment will not affect other communication between end-users. Also, this structure is easy to expand and move.
T T Figure 4. Software Architecture
R Figure 2. Start network structure
A. The Overall Design of Terminal Sensor Nodes In this system, the different terminal nodes have to collect ECG, blood oxygen, blood pressure, body temperature and other physiological signals, which have diversities in physiological characteristics. At the same time, various sensing nodes have many common features, such as LCD liquid crystal display, wireless transmission, human-machine interface. Based on this, we develop a common platform which has common features and several analog front-ends. The analog front-ends collect ECG, blood oxygen, blood pressure, body temperature and other physiological signals. The common feature platform and the analog front-ends are connected by analog front-end interface, just as shown in Figure 3. Overall, this design can achieve the goal of flexibility and versatility. Figure 5 shows our implementation of such design.
The software architecture on the terminal sensor nodes is designed in this way: application software can be found on the main.c file; from here, the main() routine can call different service and hardware features. Services will have direct contact with hardware on the MCU. The hardware has a hardware abstraction layer that eases migration to another MCU. The fact that each module has its own .c file simplifies the process of adding and eliminating modules in the code.
Figure 5. ECG monitor nodes
C. The Selection from Low Band to High Band In general, the traditional short-range wireless communication use low-frequency band, which is less than 1G ISM. Low-frequency means that a low data rate, so that equipment will work most of the time, rather than sleep, thereby increase the power consumption. Take wireless mouse for example: RF data rate: 9.6kbps
Figure 3. Common Platform and AFE
Packet length: 100bits Typical mouse up-date rate: 10-15 ms To transfer one package will take:
B. The Software Design of Terminal Sensor Nodes The software design pursues these goals: ? Modality – The software must be completely modular and with as little cohesion as possible. This modularity should be reflected in ease if making changes, if adding
100 bit 9600 bps
A new package from the mouse is in other words ready for transfer as soon as the last one is finished. Since the system has to stay on the air for as long as it is in use, the system’s power consumption is relatively high. If using the 2.4G high-frequency band, the present industrial level can make the RF transmitter’s data rate over 1Mbps, thus the above-mentioned application can transfer a package in 100us, as show below:
100 bit Ttr = 1 Mbps = 100us
communication may well be able to operate in this region with only some degradation of system sensitivity as a result.
Our system is lithium battery powered, which requires very low power consumption. That is the reason we choose 2.4G band for short-distance wireless communication. To sum up, 2.4G frequency band has the following changes over <1G band: <1G: low data rate = long time on air = collision likely = strict frequency separation needed 2.4G: high data rate = short time on air = collision unlikely = time sharing on one frequency partly replaces the need for the strict frequency separation. Based on the above analysis, the system’s short-range communication use low power 2.4G wireless transceiver chip nRF2401. D. Co-existing with Other Systems To make a robust system, it is important to understand how other systems operating in the 2.4GHz bands affect your system. There are about 4 categories: ? Direct spread spectrum devices(WLAN) ? ? ? Burst and frequency jumping systems (nRF24xx and Bluetooth). Single channel, low data rate systems. Multiple systems of same type(Multiple nRF24xx systems)
Figure 7. nRF24xx operating at outer limits of WLAN spectrum
WIRELESS COMMUNICATION PROTOCOL STACK
The system’s wireless sensor network topology is start network, as shown in Figure 2. nRF24xx uses a proprietary protocol rather than standard protocol like Bluetooth or Zigbee. This means simplicity and flexibility. We can implement out own communication protocol based on our requirements. The architecture of our own wireless communication protocol stack is shown in Figure 7.
Figure 8. Wireless communication protocol stack architecture
Since WLAN is widely used in daily life, here we take WLAN for example to analyze the co-existing of nRF24xx and WLAN.
As can be seen from Figure 8, this protocol stack is divided into three layers: the physical link layer, network layer and application layer. Link layer mainly deals with chip-related settings. Network layer mainly deal with node’s joining, existing, associated management and provides init, ping, listen, send/receive, I/O and other functions. The application connected to network layer through port number, similar to TCP/IP protocol stack. Typical applications have predefined port and customer applications have the unreserved ports. V. HOME GATEWAY
Figure 6. nRF24xx co-existing with WLAN
It can be seen in figure 6 that typically, only one WLAN network will physically co-exist in any particular area. This means that 75% of the ISM band is WLAN free, this equates to 60+ channels free for nRF24xx operation. Figure 7 shows that at the outer channels utilized by the WLAN transmission there is a sharp fall-off in signal strength and therefore the nRF24xx
Home gateway requires relatively strong processing power, storage capacity and communication capacity. It connects wireless sensor network and Internet to enable communication between two kinds of protocol stack. It works in this way: monitoring nodes, collecting signal and forwarding them to the Internet. It is both an enhancement of the sensor nodes, which have sufficient energy supply and more memory and computing resources, but also a special gateway, which has a wireless communication interface but does not have monitor
function. Based on the above mentioned considerations we have adopted a Samsung ARM9 chip (S3C2440) for the control chips and Linux for the operating system to design our home gateway. We use QT platform for GUI system. It provides powerful features for user to construct graphical interface. QT also provides a rich multi-thread programming support. Each thread has different functions and tasks, and carried out synchronization through QT synchronization mechanisms, such as semaphores, synchronization locks, etc. We divide the specific function modules into the following diagram, figure 9.
Figure 10. Signal processing flow
Figure 9. Home gateway modules
Home gateway's signal processing flow diagram is shown in Figure 10. It receives physiological data through wireless module which is SPI nRF2401. The nRF driver, which is character device driver , provides a mechanism for user space access. Demultiplexer module is in the user space. By the standard IO interface, read(), it reads mixed collected sensor signals and sensor parameter, then store the received data in buffers for the corresponding signal processing modules. Signal processing module now includes pre-processing module, heart rate calculation module and ventricular tachycardia and ventricular fibrillation detection module. if abnormal heart rate, blood pressure, blood oxygen, ventricular fibrillation, or ventricular tachycardia is detected, signal processing module will send message to the alarm module. Alarm module will use TCP/IP protocol via Ethernet driver to send alarm message and the related data to Central station. In addition, the output of the signal processing module via SD driver will store into SD memory card for as long as five day, allowing future analysis. Meanwhile, the signal waveform is shown in GUI interface in real time.
Using this architecture (sensor monitor – home gateway – central station), features which require relatively more energy consumption, computing resources, and fixed connections can be located in home gateway. Features which target at lowpower portable applications can be located in the sensor terminals. In this design, because home gateway has strong computing resources – ARM processor, and fixed power supply, more complex physiological signal processing and anomaly detection can be enforced in home gateway, only sending abnormal signal and alarm information to central station. This not only can significantly reduce physiological signal transmission network bandwidth requirements, but also can off-load intelligent diagnosis work to different network nodes, reducing central station’s analytical processing workload as a whole. On the other hand, as home gateway is distributed in the patient’s home, monitoring program can be customized for the patient’s condition, leading to more efficiency of diagnosis. Sending alarm message from central station to patient’s family, 120, or community doctors via GPRS, on one hand can reduce time for patient’s relief, one the other hand can help doctors grasp more comprehensive condition of the patient and get support from a large hospital doctors. As a whole, this can improve the success rate of rescue greatly. Obviously, this process is unable to complete without establishing a linkage mechanism between medical institutions, patients’ family and other social resources. This remote healthcare monitoring system is collaborated with Sun Yat-sen Institute for cardiopulmonary-cerebral resuscitation, including body wearable monitoring monitors, wireless home gateway and center station. The wearable monitor can collect ECG, respiration, non-invasive blood pressure, and oxygen saturation signals identify these signals and alarm when abnormal happens. The home gateway can save signals data up to five days and do signal processing like
filter signals. At present the entire system has been built basically completed, but there are places where some details are now being perfected, such system reliability and scalability. In future, more function terminal nodes will be added to the entire system. Intelligent diagnosis algorithm will also be developed to help doctors. ACKNOWLEDGMENT This work is funded by Chinese High Technology Research (863) Funds (Funding No: 2007AA01Z308), the National Natural Science Foundation of China (60,904,031), the CAS Knowledge Innovation Program and Xiamen Science and Technology Project (3502Z20093005) Acknowledgements are given to Prof. Hu Cao, and Prof. Liao Jingshen, who helped me and gave me constructed suggestions. REFERENCES
 Fulford-Jones TR, Wei GY, Welsh M, A Portable, Low-Power, Wireless Two-Lead EKG System, 26th IEEE EMBS Annual International Conference, September, 2004.J. Clerk Maxwell, A Treatise on Electricity and Magnetism, 3rd ed., vol. 2. Oxford: Clarendon, 1892, pp.68–73. Philip L, Sam M, Joseph P, Robert S, Kamin W, Alec W, David G, Jason H, Matt W, Eric B, David C, TinyOS: An Operating System for Sensor Networks, Ambient Intelligence, Springer, 2005 B. Zhou, C. Hu, HB Wang, RW Guo, and Max Meng, "Wireless Sensor Network for Pervasive Medical Supervision", Proceedings of the 2007 IEEE international Conference on Integration Technology, 740-744, 2007 Shu-Di Bao ， Yuan-Ting Zhang . A design proposal of security architecture for medical body sensor networks. International Workshop on Wearable and Implantable Body Sensor Networks, April 2006 Blanco-Velasco M, Cruz-Roldan F, Lopez-Ferreras F, A low computational complexity algorithm for ECG signal compression, MEDICAL ENGINEERING & PHYSICS, 26(7): 553-568, 2004 TI company, Introduction to SimpliciTI Nordic company, 2_Ghz Link Integrity Thomas HC, Charles E, Introduction to Algorithms, the MIT Press, 2001