Wearable computer as a multi-parametric monitor for physiological signals

Wearable Computer as a Multi-parametric Monitor for Physiological Signals Julio C. D. Conway Antonio 0. Fernandes

Claudionor J. N. Coelho Jr. Luis C. G. Andrade

Diogenes C. da Silva. Jr. Hervaldo S. Carvalho

Computer Science Dept. - UFMG Federal U. of Minas Gerais, Brazil {conway,coelho,diogenes,otavio,lcga,hervaldo} @dcc.ufmg .br

must acquire the physiological signals, process and send the signals through a computer network, while respecting the time constraints. The main characteristic is its wearability, i.e. provide free motion and location independence to individuals. Due to its flexibility and high-speed, this equipment can be used inside hospital facilities allowing patients to move while being monitored without requiring the patients to carry or pull heavy equipments. It allows patients to be monitored from home continuously even when performing normal activities. Finally it allows patients to freely perform outdoor activities in regions with wireless network coverage. This paper is outlined as follows. Some background of medical systems, mobile computing and wearable computers are given in Section 2. Section 3 discusses the clothing and usability aspects of the equipment. Section 4 describes the hardware, operating system, power management and networking of the device, followed by conclusion and future directions.

Abstract This paper describes the design of a wearable multiparametric physiological signal moniror, a continuously running internetworking system for monitoring multiples physiological parameters of individuals during daily tasks. A pi.otot_vpe, measuring parameters such as electrocardiogram, oximetiy and non-invasive blood pressure is now in test. The system peiforms real-time processing on the signals and can send these parameters to a central monitoring station over a computer network. The relevant points of the control operating system and applications are described, along with future directions for research.

1. Introduction The high costs of hospitals treatments and the necessity of home care assistance brought an increasing demand of the auto patient monitoring. Until recently, fixed medical equipment requires a large bandwidth to transfer data to a monitoring station, while a mobile medical equipment (used in telemetry) [3] is able to handle only a small number of parameters to be monitored at lower quality (due either to the reduced bandwidth to communicate to a monitoring station or to the limited storage in such devices). The continuous reduction in component sizes of increasing performance and complexity, associated with high-speed wireless networks that are being installed in several facilities and major urban areas makes it possible today attach to a mobile person a computer system with high performance and that communicates with the world at high-speeds. This computer system has been named in the literature as a wearable computer [21. The objective of this work is use a wearable computer with wireless networking capabilities for monitoring multi-parametric physiological data. This application

2. Background 2.1 Computers in Medicine The monitoring area of physiological signals has grown in recent years, changing from a simple one parameter monitor to a truly workstation, where several signal are shown and multiple analysis of these parameters are done, allowing physicians to make complex studies and give secure diagnostics. Included in such parameters are EKG', oximetrf, non-invasive and invasive blood pressure, temperature, cardiac debit and EEG3 [9]. Eletrocardiogram

* Blood oxigen level

' Electroencefalogrdm 236

0-7695-0862-6lOO $10.00 0 2000 IEEE

characteristic is their small size with a high processing power similar to desktops. This trend has allowed new applications of embedded systems that are compact, powerful and with connectivity.

Sensors are used to pick up the electrical activity of the human body, such as electrical activities and movements. There are sensors suitable to all kind of parameters, and they can be either external or embedded. Currently, these monitors are systems based on microprocessors or microcontrollers, responsible for handling the analog-to-digital conversion of signals, processing, storing, printing and visualization of signals. An important function, mainly nowadays, is the ability to send these signals to a remote location through a computer network. With this characteristic, the monitor can be connected to a local network inside the hospital, where a central point can monitor several patients at the same time. A remote patient, maybe in another country, can be plugged to this network, and can be monitored at the same way. This characteristic is specially important for the so called “second opinion”, where one or several specialists can visualize the signals of a patient and send their opinion about the case to help the physician taking care of the patient. It is an agreement that a patient in a pos-surgery situation improve the health doing some activities, like walking on the hospital’s garden, sitting on a television room or even staying at home in recuperation. As a result, fixed equipment reduces or limits considerably the mobility of such patients. On the other hand, most current mobile devices lack the computing power and the network bandwidth to completely monitor a patient in such conditions. So, it is imperative that the monitor must have wireless networking capabilities to allow the patient to move freely while the physiological parameters are continuously monitored and sent to the Central Station.

2.3 Wearable Computers Another area that is emerging is wearable computer, a computer that an individual can wear like a cloth, and can input commands and run applications while walk or do some daily activity. This kind of computer is integrated in the personal space of the user, controlled by the user and is always operational and available. [ 11 Unlike conventional computers, a wearable computer runs the application continuously and is always ready to interact with the user. We can define more specifically a wearable computer as a computer that is operational while portable and allows hands-free use. Besides, a wearable computer must have sensors to interface the physical environment. It must be proactive, in the sense that it must send information to the wearer even when this information is not been required. Finally, the wearable computer must be always on, always running. This leads to a new problem, namely, the power consumption. Power management arises as a main optimization problem found in wearable computers. Section 4 explains the techniques used in the monitor for optimization of power consumption. [8]

3. Clothing An important issue in wearable computers is wearability. Wearability is defined as the interaction between the human body and the wearable object [I]. The main goal is to define spaces on the human body where solid and flexible forms can rest without interfering with fluid human movements. The dynamic nature of the human body, movements and shapes should be observed when designing the garment. There are some design guidelines for wearability [ l ] that were adopted in the design of the monitor. In the following paragraphs some of these guidelines are presented. First, a wearable device requires unobtrusive placement. Criteria for placement can vary with the needs for functionality and accessibility. Thus, we must identify areas of the human body suitable for placement, for example areas that have relatively the same size among adults, areas that have low movement but flexible even when the body is in motion and large areas. Thus, the main areas for suitable placement in the human body [ I ] are the collar area, the rear of the upper arm, the

2.2 Mobile Computing The mobile computing arises as a new computational paradigm, providing continuous connectivity to a computer network, independent of physical location. The technological evolution provides access to information anywhere in the world. Satellites can cover the entire globe, and with the new wireless communication technologies, old barriers such as bandwidth, distance and geography are eliminated [lo]. The recent years has shown an explosive growth in the number of laptops and notebooks sold, and the number of nodes connected in the Internet. Another growth is in the market of wireless devices with networking capabilities. Thus, investments were made in the wireless communication devices and the support to the devices. At the same time, there is a new generation of embedded CPUs arising in the market. Its main

237

compatibility with the hardware and software architecture of PC based systems in the form of stacked modules of 3.6 x 3.8 inches, providing a practical way to implement a modular PC based architecture for embedded systems. With this architecture in mind, we selected the DIMM-PC/486-I, a PC104 from JUMPtec, that integrates into a motherboard the functionality of a 80486 DXlOO CPU, System-BIOS, 16 Mbytes DRAM, keyboard controller and real-time clock, but instead of a silicon disk it has a 16MB IDE compatible flash hard disk onboard. Integrated on board, the DIMM-PC/486-1 also offers additional peripheral functions like COMl , COM2, LPTl, floppy interface and IDE-hard disk interface. Moreover, this system and traditional PC are electrically compatible.

forearm, the rear, side and front ribcage, waist and hips, thigh, shin and top of the foot. Another important point is the physical form of the wearable. It must accept human convexities, and thus, must form a concavity on the inside surface touching de body. Aside the natural human body convexities, the wearable must allow the flexing and extending movements of muscle and tendons beneath the skin. Also the form of the electronic hardware must be taken into account when designing a wearable computer. A well suitable place to put a wearable is the torso, but the arms need to have full freedom. In addition, the torso needs the full ability to twist and bend. The design must consider the great variation of the torso size among several individuals. Thus, the design must consider these variations to allow as many types of users as possible. The wearable’s weight must not hinder the user movement. Thus it is essential that the weight must be kept as small as possible. In the human body, the waist and hip areas support most of the human extra weight. Thus, it is suitable to put the load there. The correct place of the sensors is a special issue to design. With these wearability considerations in mind, we designed a wearable computer to accommodate the hardware and associate and intended to be a comfortable cloth to the user, yet a complete physiological acquisition system.

4.2 System Description In this application, the individuals are patients and the system is used to collect the physiological signals related to patient’s health. However, this system can be used for any type of application where remote monitoring through computer network, such the Internet, is desired. The system is composed of a wearable module and a remote module. The wearable module consists of a physiological signals acquisition module and the processing, storage and communication module. The remote module can be any computer connected the same network of the monitor. This remote module is in fact the remote central monitoring, where the physiological parameters such as the EKG, oximetry, pressure and the associate traces can be displayed in a graphical screen. Figure 1 shows a block diagram of the system.

4. System Architecture We designed a wearable medical equipment to work on mobile environments to be worn by a subject as

clothes. The device provides location independence and minimum reduction in mobility. We describe the main aspects of the hardware and software architectures below.

4.1 Platform Sensors

,

1

Storage Communication

7

Remote Central

Processing

In order to simplify the design and decrease the delay of the development process we choose to use available standard hardware and software modules, instead of designing from scratch. We chose a computer architecture based on the Intel x86 to provide the desirable standard hardware and software. Among the options the PC104 standard [6] provides a modular approach to embedded systems. The modules are designed to provide a great variety of functions for embedded systems, including fixed, portable and mobile environment. PC 104 redefines the traditional Personal Computer (PC) enclosure dimensions, while satisfying the requirements of strength, reliability and length constraints of embedded systems. It offer full

-+ phone

+ mobile

-+

wireless

The sensors acquire the patient physiological signals and are specialized for each monitoring parameter. These signals are sent to the acquisition module. The acquisition module sends the samples of each parameter to the processing, storage and communication module, where the samples can be stored or sent through a 238

computer network to the central monitoring station. This transmission can use conventional telephone lines or mobile phone through the use of a modem or can use a wireless data communication link. At the other side, the central can send commands to the monitor. These commands can act on the patient passively (orientations to be followed, through messages on a LCD display and audible signals), or actively (the monitor can, for example, control the injection of a drug to the patient). After the acquisition of the physiological signals, the acquisition module amplifies the signals to raise the voltage level to suitable values for processing. Next, the signals are filtered, removing undesirable noises. After that, this treated signal is digitized, through an ADC (analog-to-digital converter) and available in the form of byte samples [4] [ 5 ] . Finally, these byte samples are sent to a microcontroller. This microcontroller has the task of controlling the ADC and sending the samples to the processing, storage and communication module. The interface between the aquisition module and communication module are optically isolated, isolating the patient from the AC power (if the battery is not used) and the transmission system. Figure 2 shows a complete view of the system.

PCMCIA controller card, allowing the wireless connection of the monitor with a fixed network. This module receives the samples from the acquisition module, stores the samples in non-volatile memory and sends it to remote monitoring central. The Input/Output interfaces of the DIMM-PC/486-I connect the processing module with the acquisition module, through the parallel interface, and can be used to communicate with other peripherals, like a printer. The PCMCIA controller allows the use of wireless data transfer. The wearable system uses a rechargeable battery, although it can also use the AC line. The monitor is a wearable device, comfortable to the user, without loss of movement. Nevertheless, the device can also be used as a desktop monitor. The complete system, modules, battery and sensors, are fitted in a specialized garment, according to the specifications of section 3. For the wearable device operation, the individual must wear a special cloth, place the sensors and turn the monitor on. After that, the system automatically begins to acquire the physiological parameters, boot the operating system, start the communications and connect to the remote monitoring station. From this point, the multiple parameters are stored and continuosly sent to the remote station. '

High-speed wireless link

4.3 Software Specification 4.3.1 Remote Module Specification The remote module is composed by the following software components:

Remote HTTP Client: any computer connected in Central

1 ,

the network with a HTTP browser. This client requests the samples and values of the parameters from the wearable monitor HTTP server.

Wired link

Remote TCP/IP Client: any computer connected in the network with a TCP/IP stack. This client requests the samples and values of the parameters from the wearable monitor TCP/IP server.

Remote Graphical Station: any computer with Microsoft Windows 95 or higher, connected to the Remote HTTP Client or the Remote TCP/IP Client. This module displays the parameters values and associated graphical curves at the screen.

U

Figure 2. Complete view of the system

Remote Parameters File server: any computer with Microsoft Windows 95 or higher, connected to the Remote HTTP Client or the Remote TCP/IP Client. This module stores the parameters received by these clients and provide this data to the Remote Graphical

The processing, storage and communication module consists of a DIMM-PC/486-I (described in section 4.1). It runs a Linux Operating System [7], allowing the utilization of the TCP/IF' stack. This module has a

239

until values quite bellow of the nominal frequency of the processor, or put the processor in the “sleep mode”, allowing just memory to be energized. We can also put the communication channel in the inactive state. Adding adaptability characteristics, the communication assume special characteristics, making the communication protocol more complex, leading to a message exchange between the monitor and the central.

Station or any application that wishes to use the parameters data.

4.3.2 Wearable Module Specification The wearable module is composed by the following software components: 0 Acquisition Module: responsible for acquiring and formatting the data from the sensors and to send this data to the HTTP or TCP/IP physiological parameters servers. D Physiological Parameters HTTP Server: servers the HTTP requisitions Erom Remote HTTP Clients. WEB clients all over the world can have access to the patient information, since they are connected at the same network of the monitor. 0 Physiological Parameters TCP/IP Server: servers the TCP/IP requisitions from Remote TCP/IP Clients. TCP/IP clients all over the world can have access to the patient information, since they are connected to the same network of the monitor.

A critical factor is the battery load. The microcontroller frequently monitors the battery load, and the firmware can change some communication parameters to adapt to any change in load battery. Thus, the number of physiological parameters can be gradually reduced when the battery load is decreasing in the 90% to 70% range below its nominal value. As the load reduces, say between 60% and 70%, the firmware can slow down the clock of the CPU, in order to reduce power consumption, although transmitting. With the load below 60%, the firmware puts the system in power-down mode, that is, the processing and transmission stop, keeping just the data memory with a minimum power. The energy failure can be reported to the remote monitoring station as a procedure before entering the power down mode. Another parameter that affects the communication is the error rate, which is periodically checked by the remote station. When this rate grows to unacceptable values, the remote station requires that the monitor to reduce the transmission rate. Similar procedure is taken when the signal-to-noise rate is high, providing mechanism for adapting to channel conditions. At the communication protocol level, the physiological signals are sent from the monitor to the remote monitoring station. Usually, the samples are acquired with a resolution of 12 bits and a sampling rate of 1 KHz. This is the normal operation of the system. Again, according to the communications channel conditions and the battery load, the parameters can be sampled at lower rates, such as 600 or 300 KHz. This reduction in the sampling rate leads to fewer bits per second to be transmitted; demanding lower load from the system. Summarizing, the design of the software of the multiparametric wearable monitor takes into account dynamic power management aspects outlined above, allowing an acceptable performance in power consumption to this particular application.

4.4 Power Management A critical factor on the performance of mobile systems is power consumption, because generally batteries power these systems. The main goal is delivering high performance with a limited consumption of power. High performance is required by the increasingly complex applications, sometimes multimedia applications running in currently portable devices. Low power consumption is necessary to produce acceptable autonomy in batterypowered systems. Thus, to produce mobile computer systems with highly energy-efficient is a great challenge in electronic design [8]. To be competitive, an electronic design must be able to deliver peak performance when required. Nevertheless, peak performance is required just in some time intervals. Similarly, some part components are not always required to be in the active state all the time. The capacity to enable and disable components, to adapt to environment conditions, as well as to steer performance according to user requests are fundamental to produce energy-efficient designs. Thus, a new concept arises: the dynamic power management. It is a methodology that dynamically reconfigures an electronic system to provide the requested services and performance levels with a minimum number of active components or a minimum load on such components [8].

4.5 Current state of the prototype Currently the wearable part of the prototype uses a DIMM-PC/486-I with a PCMCIA controller card for wireless communication, an acquisition card and a

To achieve this goal, we can selectively turn off or reduce the performance of the components when they are idle. Specifically, we can reduce the clock of the system 240

battery of 4 ampshour. The garment is based on a bag tied at the waist through a leather belt.

performance, dynamic power management techniques must be used to allow the system to provide the require service within acceptable performance levels and power autonomy.

The Central Station is a MS Windows based computer connected to the same LAN (local area network) that the wearable monitor. The running application shows the beats per minute at the Central Station window. The current prototype uses a wireless data link to tranfer data between the monitor and the remote Central Station.

6. Future work In the current stage, the monitor prototype sends just the EKG through a wireless link, using wireless Ethernet PCMCIA adapter card (WaveLan) from Lucent Technologies and the WavePoint device associated, connected to a LAN. The next step is to add other physiological parameters, namely NIBP and SA02. The monitor then will send to the remote monitoring station the EKG signal trace, the NIBP values and the S A 0 2 trace and values. These values and traces will be shown in a graphical screen at the monitoring station.

Tests were made and an individual wearing the prototype walks around 50 meters for approximately 2 hours, and the EKG was continuously transmitted and displayed at the Central Station. The monitor was tested using a single cell wireless LAN network of 2Mbps. In a multi-cell network the communication coverage would have been increased. In metropolitan areas of the US, there is currently availability of RF wireless networks with low rates, e.g. 9.6 or 14.4 Kbps [Ill. Higher capacity metropolitan wireless networks are planned with rates up to 512 Kbps. We are now trying to optimize the battery performance. Our goal is to reduce the size and increase the time cycle. For this, management techniques are being applied to reduce power consumption.

Another parameter that can be implemented is the EEG. The wearable monitor measure this parameter while the patient can walk freely using a mounted head sensor while measures of brain activity are sent to the monitoring station. Also sleep research can use the monitor, mainly for people with somnambulism. This system can evolve to a complete networking monitoring system, with TCP/IP and HTTP parameters servers that can be used for real time monitoring and analysis. A wearable computer provides a computational power and communication bandwidth larger than used in the current monitor. More functionality can be added without comprising its performance. Textual communication, such as instructions to the patient, commands to activate specific units in the monitor and voice communications are currently under development.

Wearability studies also have been done to optimize de garment in order to provide more comfort and freedom to the wearer.

5. Conclusions A mobile system has special characteristics where the dynamic changes of the environment and the adaptive capacity of the system are essential to the performance. Several applications arise from the use of wearable computes. Applications in medicine arise as a challenging field of research. Hospitals have invested a considerable effort in new technologies, allowing home care and outdoor activities monitoring. The multi-parametric wearable monitor arises as a solution for these cases where it is desirable to remotely monitor individuals, allowing mobility and connectivity through a computer network. Due to its flexibility and high-speed, patients are able to move while being monitored without requiring the patients to carry or pull heavy equipments. They can be monitored continuously even when performing daily activities. It can be even performed in regions with wireless network coverage. Power optimization is an essential factor to allow the system to achieve the required performance. The battery technology has evolved slowly, compared with the hardware technology. To compensate this low battery

References F. Gemperle, C. Kasabach, J. Stivoric, M. Bauer and R. Martin, “Design of Wearabilily”, 2”d.Intl. Symposium on

Wearable Computers, ISWC 98. Pittsburgh, PA, 1998. S. Mann, “Wearable Computing as means for personal empowerment”, Keynote address, ICWC 98, 1998. http://wearcomp.org/wearcompdef. html J. D. Bronzino, “Biomedical Engineering”, Trinity College, Hartford, Connecticut, CRC PRESS, 1995. W. J. Tompkins, “Biomedical Digital Signal Processing”, University of Wisconsin-Madison, Prentice Hall, NJ, chapters 1, pp. 1-5, 1995. C. Marven and G. Ewers, “A Simple Approach to Digital Signal Processing”, Willey-Interscience, New York, NY, chapters 2-4, 1996. R. Lehrbaum, “Using PC/104 Embedded-PCs in Mobile and Portable Applications”, AMPRO Computers, Sunnyvale, CA, pp. 1-10, 1997. 24 1

[7] Linux On-line (http://www.linux.org) and Embeddable Linux Project (htto://www.linuxrouter.com). [8] L. Benini, A. Bogliolo and G. De Micheli, “A Survey of Design Techniques for System-Level Dynamic Power Management”, IEEE T-VLSI, June 2000, pp. 299-316. [9] Nenov VI, Buxey F., Yamaguchi Y., “BRAVO/TeleTrend: A Comprehensive WWW-based neuromonitoring system for the neurosurgery ICU”, Stud Heath Techno1 Inform 1999, 62~228-34 [lo] Perkins, Charles E., “Mobile IP”, IEEE Communications Magazine, May 1997, pp.84-99. [ l l ] Comeford, Richard, “Handhelds duke it out for the Internet”, IEEE Spectrum, August 2000, pp. 35-41.

242