and some technical hints ... to follow up elderly people a home, is a particular field of application ..... through equipment lending and technical support : Atral s.a..
The Telemedicine Home Care Station : a model and some technical hints Norbert Noury1, Vincent Rialle1, Gilles Virone1
Abstract- This paper deals with the central element of the remote monitoring system for patients at Home: the telemedecine home care station (THCS). Following more than 10 years on the subject and several technical implementations, the authors, in an effort of modelisation, propose an architecture of the THCS with some technical solutions.
I. INTRODUCTION The remote monitoring of patients, for chronic diseases or to follow up elderly people a home, is a particular field of application and development inside the promising, rapidly growing up industry of telemedicine. The basic model for the telecare process looks like a « control loop » (Fig. 1) with two data flows [1] : HOSPITAL PRACTITIONER
GENERAL PRACTITIONER
PATIENT
DIAGNOSTIC
Therapeutic Protocol Therapeutic Project
+
+
Medical File
Physiological Descriptors
Patient Status HOSPITAL AREA
TELEMATICS
DATA BASE
-
TELEMATICS
HOME
Fig. 1. A model for the telecare process
In this process, the hospital practitioner is still the manager of the therapeutic project. But with the distance, he delegates the implementation to the general practitioner with the help of the « field workers ». In this information system, the forward chain carries out the therapeutic protocol (i.e. prescription, medication, etc) and the reverse chain brings back the information about the patient status (physiological descriptors, indications, etc). All these information is gathered in a data base, which may as well be distributed, for a later use. The follow up of patients at home must satisfy the same safety standard as it does in the hospital. The architecture of this information system can be achieved with standard bricks, with the help of specifically designed intelligent assistants. One of the main brick of the architecture is the home care station, a communication device which is placed in the TIMC-IMAG, UMR CNRS 5525 38706 La Tronche, France
home to facilitate the networking of data in between the patient at home and the different people in charge of his health and security. The Home care station not only plays the role of the human to machine interface, but it also brings the following features : - a local data storage (memory), - a local processing facility (real time alarm agents), - a communication mediator, - an authentication agent (PIN), - a “fire wall” to preserve data from piracy. For some reasons, a centralized architecture of the home care station is preferred to a distributed one. II. ELEMENTS OF THE MODEL Following earlier work on the home care telematics project [1], we had defined this device from the point of view of his different actors and of their particular needs. A. The actors of the system There are seven profiles or "classes" of persons likely to use the system : • the patient , • the physician (the general practitioner as well as the medical specialist), • the medical workers team (nurse, physiotherapist, nursing auxiliary), • the social workers (home help), • the emergency staff (warden, physician), • the casual operator (relatives, neighbor, helper), • the technical engineer (installation, configuration, maintenance) B. Data handled by the system. The medical file is made of several information : administrative data (name, address), references of the medical care centers, successive prescriptions, the complete record of descriptors. This may well be completed with biological results and images. The prescription, usually written by the hospital physician, contains the medicament dosage together with the descriptors acquisition procedure.
The descriptors are the data on the status of the patient : physiological descriptors, physiological signals, activity scores, etc. The electronic mail is made of textual data the different actors can send to each other. The identifiers : each actor has a personal identification number (PIN) which holds his rights of access to the information, the services. It also permits to monitor the usage history for each actor. C. The descriptors and associated sensors The descriptors gathered in the Home care station are not limited in kind or number. They fall into three categories depending on the information they handle : - the physiological sensors carry the information on the medical and biological descriptors of the person. Those can be signals as well as unary data. - the activity sensors deliver information from localization or movement sensors. The localization of the person in the home (in the home referential) allows to obtain the motor activity. From movement sensors "on board" the person (in the person referential) we get the information on the posture (standing, lying), the movement (walk), and even the fall. - the ambient sensors (temperature, pressure, hygrometry, illumination) inform on the measuring conditions, but can also directly affect the patient status (a lack of illumination can lead to morbidity or to a fall). D. The arrangement of the sensors The location of the sensors directly determines the data gathering methods. We found 3 kind of arrangements : - the fixed sensors corresponding to a sedentary gathering. The person must come to the sensor for the measure to be performed. The sampling frequency is low. - the movable sensors are used in different possible location in home. The sampling frequency is low to medium. - the ambulatory sensors, on board the person, allow higher sampling frequencies. For instance physiological signals can be monitored. We might as well make a distinction from the way the measure is transmitted, either automatically diffused by the sensor or on demand from the home care station. E . Some examples of descriptors and their sensors. Some of the usual descriptors follow, with their sensors : - Arterial Pressure (Sphygmo manometer, Wrist Tonometer) - Cardiac Frequency (Electronic Stethoscope , ECG) - Pulse (Optical pulsemeter, Extensometric Belt ) - Breathing Frequency (Plethysmographic belt) - Blood oxygenation (Oxymeter) - Weight (Weight Machine) - Temperature (auricular Thermometer) - Presence (Infra-Red Volumetric) - Passage (Door contact) - Fall (Accelerometer) - Physiological Vibrations (Mechanical Vibrations) - Posture (Angle sensor) - Smoke (Smoke detector) - Gas (CO detector)
- ambiant temperature (Thermometer) - Hygrometry (Hygrometer) - Atmospheric Pressure (Barometer) - Illumination (Luxmeter) - Sounds (Sonometer, microphone) F. Functionality to be implemented For each class of actor, we can define which functionality they expect to find in the system. The patient: Authentication (PIN number) Read his medical file (the accessible public part) Read prescription Read or write his descriptors Read or write a message in mailbox The physician: Authentication (PIN number) Read the medical file (public and private part) Read prescription Read or write the descriptors Read or write a message in his mailbox The medical worker: Authentication (PIN number) Read the medical file (public part) Read prescription Read or write the descriptors Write an intervention report Read or write a message in his mailbox The social worker: Authentication (PIN number) Write a visit report Read or write a message in his mailbox The emergency staff: Authentication (PIN number) Read the medical file (public and private part) Read prescription Read the descriptors Write an intervention report The casual operator: Authentication (PIN number) Read the medical file (public part) Read prescription Read the descriptors Read or write a message in his mailbox The technical engineer: Authentication (PIN number) Read the software version Load a new software version Adjust the local tuning parameters III. ARCHITECTURE The architecture proposed can be compared to the human nervous system physiology : - the central brain is directly in charge of the important perception functions and supervises the others major functions : this is the visible part of the home care station with its user-interface and the communication facilities,
- the central nervous system is the spine of the network : this is the “backbone” of a local “field bus” (Local Area Network – LAN), - the peripheral nervous system connects the effectors and affectors, but also deals with the “reflex” functions : this is the wired and wireless interface with the sensors and the actuators, with a local decision (first level alarm generation). The local field bus [2] is a mediator that simplifies the intercommunication in between the information devices, producers and users, that are spread over a geographically limited area, here the flat. The field bus covers 3 mandatory levels of the OSI model [3]: - the level 1 (Physical layer - OSI level 1), defines the mechanical and electrical access to the physical medium, as well as the data coding , - the level 2 (Link layer – OSI level 2) controls the access and share of the medium (Medium access control - MAC) , and handles the errors (Logical Link Control – LLC). - in the level 7 (Application layer - OSI level 7), high level features are implemented such as the addition or withdrawal of a communicating device. Additional layers may be implemented to resolve the addresses (Network layer - OSI level 3). For the sake of flexibility and to cope with the heterogeneity of data gathering, two kind of field bus can be set simultaneously or not depending on the situation : wired or wireless bus. A The wired local bus The wired local bus can be easily implemented through : - the existing electrical network after some minimal adaptations (geographical redistribution in every rooms , isolation from the outer network for confidentiality) - a single pair of telephone wire spread all over the flat with plugs fitted in every rooms. A second pair of wires can transport a direct current power supply for those sensors which doesn't hold a battery or need to recharge it. In any case the wired bus is reserved for the fixed or moveable sensors. It must support a multi-master access endowed with a collision detection algorithm. B The wireless local bus Two main techniques can be considered for short range wireless communications in the home : - The optical link (Infra Red) requires an unrestricted, lineof-sight, view in order to communicate effectively between devices. This is of course a real limitation for the free motion of the person around the home. - The RF link enables non-line-of-sight communication and better suits an indoor mobile task. The choice for the carrier frequency naturally falls to one of the authorised one. The emitting power should be chosen so that to reduce the environment pollution and reduce the covering area to the flat in order to preserve data confidentiality.
C Physical Limitations of RF links There are 2 physical effects to be taken into account for a data communication with a "human mobile" in a building [4]. The first one, the Doppler effect, induces a frequency shift due to the motion of the transmitter embarked on the person. It is negligible (about 3 Hz) for a maximum displacement speed of 2 meter per second, and a 433 MHz carrier band. The second one is the Rayleigh fading [5] which introduces an amplitude variation of the electrical field with space due to several interference with reflected waves. This phenomenon introduces blanks in the transmission which can be disastrous for any digital link. The effect is about twice the Doppler shift frequency and increases if the mobile remains in the blank region. There are two ways to reduce this effect : The first, Spatial Diversity, consists in receiving the same signals on multiple antennas spatially distributed. The cost of such a solution increases with the multiplication of the receivers. The second, Frequency Diversity, consists in using multiple carriers [6] which leads to different spatial repartition of the Rayleigh fading issue. We usually have two options to implement this strategy: - the carrier frequency is changed only when a transmission error occurs, the Multiple Channel option. - the carrier frequency is changed at a regular pace, whatever the propagation conditions, the Frequency Hopping Spread Spectrum option (FHSS). D Communication Bandwidth The nominal data rate is proportional to the message length, the periodicity of emission, and the number of sensors : R=N*L*P R, Data Rate in Bits per second (bps) N, number of sensors L, Message length (number of bits) P, Rate of Emission (Hertz) In our case, the messages are short (10-100 bits per message), the rate of emission is low (0,5-1 Hz), and the average arrangement is made of 20 sensors, then it reaches a 2000 bps data rate, closed to the 2400 bps standard rate. E The wireless data link mode The communication should be established in both directions (duplex mode), and at least in one direction at a time : half duplex mode. This is necessary to implement a "frequency diversity" algorithm to overcome the Rayleigh Fading : the maximum reception level is determined through frequency hoping and reception level measurement (RSSI). This also allows to adjust the internal parameters of the sensors or to run periodical controls on the sensor.
IV. SOME TECHNICAL HINT A The central unit The Home care station is prototyped from a PC architecture. This was a choice for the shortening of the development phase. An integrated version will be made out of a standard PC-104 architecture. B The operating system Following the former choice, we used the Microsoft Windows 98 operating system. The next version will turn to the LINUX operating system for real time operation and better security. C Development and prototyping languages The prototyping was carried out with the National Instruments LabView program, but the operational version is written in Java in order to gain the hardware independence. D Data description language The XML language was adopted as a standard for data exchange format. It is also used for the description of the data as well as the networked elements. This aims at interoperability. E User-interface The human interface is part of the "Home Care Station" but can also be placed elsewhere. In the first implementation, we used a PC computer with its classical keyboard, mouse pointer and monitor. But a further version can rely on a TV set with its remote control. F The wired local bus The Controller Area Network (CAN) [7], an industrial standard, was selected. It works on most of the mediums, one of this can be the telephone wire. It covers the MAC and LLC layers of the level 2 of the OSI model. It is a quasi real time network. Its 1 Mbps bit rate is sufficient. G The wireless local bus The choice for the carrier frequency naturally falls to one of the free ISM bands [8], 434 MHz or 868 MHz bands (European countries), with 915 MHz for the USA. It is also advantageous to use the low cost equipment and the low power consumption associated with these bands. The emitting power is limited to 10 mW which suits a flat area. The bit rate is typically 50 Kbps. V. CONCLUSIONS A Home care station has been designed to fulfill the basic need for the remote follow up of patient descriptors. This is operational but waits now to be medical evaluation. Then some other functionality will be easily added. A simple tuner PC-card will give the opportunity to display the data on the personal television screen of the person. Then a simple web cam on a USB hub, will open the possibility for televisitation. Also, the local network bandwidth has been volontarily maintained to middle rates for real time purpose. The
Ethernet technology, more costly to integrate in the sensors, could speed up the rate to 10 or 100 Mbps with acceptable delays. As for the wireless link which could benefit of the 2 Mbps of the Bluetooth technology, although it is still costly in price and energy. ACKNOWLEDGEMENTS The authors are grateful to the Action Concertée Incitative "Télémédecine et Technologies pour la Santé" of the French Ministry of Education and Research, which granted us via the TIISSAD project and consortium. We wish to thank the private companies which help us through equipment lending and technical support : Atral s.a. for the detection sensors, Atmel for the wireless RF module. We are grateful to the students Loic Boissy, Philippe Ravannat, Guillaume Danjoux, from the University Joseph Fourier Grenoble, who worked on the project during their training sessions. REFERENCES [1] N. Noury et al. , “A telematic system tool for home health care”, Int. Conf. IEEE-EMBS, Paris, 1992, part 3/7, pp 1175-1177 [2] JP. Thomesse , “Fieldbuses and interoperability”, Control Engineering practice 7, 1999, pp 81-94. [3] IEEE project 802, “Local and metropolitan Area Network standard, Draft IEEE 802.1, Overview and Architecture”, june 1983 [4] Remy JG, Cueugnet J, Siben C (1988). Systèmes de radiocommunications avec les mobiles. Ed. Eyrolles CNET-ENST. [5] Rice SO (1944). Statistical properties of sine waves plus random noise, Bell Syst Techn J, vol 23: 282-232. [6] Muammar R, Gupta S (1982). Performance of a frequency-hopped multilevel FSK spread-spectrum receiver in a Rayleigh fading and lognormal shadowing channel. ICC 1982, session B. [7] D. Paret, “Le réseau CAN, Controler Area Network”, Ed. Dunod, France, 1996. [8] N. Noury et al. , “Wireless ambulatory acquisition of high resolution physiological signals”, ETC2000, European Congress on Telemetry, Mai 2000, Garmisch Parten-Germany.
