An Environmental Multiagent Architecture for Health ... - CiteSeerX

1 downloads 0 Views 166KB Size Report
people. This paper presents a multiagent architectural solution, called .... The effort to have complete FIPA compliant messages will be little and we will ...
An Environmental Multiagent Architecture for Health Management Francesco Amigoni and Nicola Gatti Artificial Intelligence and Robotics Laboratory Dipartimento di Elettronica e Informazione Politecnico di Milano Piazza Leonardo da Vinci 32, I-20133 Milano, Italy [email protected], [email protected]

Abstract. The development of smart devices and software techniques within the fields of ubiquitous computing and ambient intelligence offers new interesting opportunities for elder, disabled, and chronically-ill people. This paper presents a multiagent architectural solution, called environmental agency, for the flexible cooperation and coordination of the devices pervading an environment in order to satisfy the needs of human beings, in particular of impaired people. We experimentally validated the environmental agency paradigm for health-related needs with special attention to the physiological monitoring of diabetic patients.

1

Introduction

The development of ubiquitous networked devices offers new opportunities for elder, disabled, and chronically-ill people [1]. The purpose of this paper is to illustrate a novel multiagent architectural solution, called environmental agency, for flexible cooperation and coordination of the devices pervading an environment in order to satisfy impaired people exigencies. The name ‘agency’ derives from our conception of a cooperative multiagent system as a single machine composed of complex components: the agents [2]. We will refer to a scenario in which two agencies are involved. Patient information acquisition is performed by an anthropic agency, introduced in [3], whose agents are placed on/in the patient’s body to monitor and regulate physiological processes. The anthropic agency can raise some exigences (alerts) that it is not able to manage; these involve mechanical and hardware issues (e.g., malfunctions of the devices) as well as biological and physiological issues (e.g., emergency state alarms and telemedicine). The exigencies are intercepted by the second agency, the environmental agency, composed of agents that are pervasively disseminated in an environment and that are able to detect and solve, among others, the requests generated by the anthropic agency. The main original contributions of this paper are summarized below. – A powerful and flexible architecture for the dynamic integration of heterogeneous devices in an unique multiagent system, based on splitting each agent in two semiagents: the operative one and the cooperative one.

– A robust problem solving technique that is dynamically and flexibly shaped on the resources currently available. – A conceptually elegant and practically effective solution to the problem of dynamically integrating networks of devices (represented as agencies) in a larger multilevel system. In our case, the anthropic agency is viewed from the environmental agency as a single agent. – The implementation of the above ideas in an experimental system in which the “bridge” between the environmental agency and the anthropic agency that regulates the glucose-insulin metabolism is provided by a palmtop computer in line with modern tendencies in health telemonitoring [4]. This paper is organized as follows. The next section briefly surveys the notion of anthropic agency. Section 3 describes the composing agents and the architecture of the environmental agency. Section 4 analyzes (both theoretically and experimentally) the interactions within the environmental agency. In Section 5, we compare our work with the current state of the art. Finally, Section 6 concludes the paper.

2

The Anthropic Agency

The anthropic agency (see [3] for a comprehensive description) is a paradigm for the development of distributed multiagent [5] regulators for human physiological processes. We implemented an anthropic agency for the regulation of the glucose-insulin metabolism in simulated diabetic patients. The metabolism of a diabetic patient is simulated starting from the work in [6] and introducing some disturbs typical of diabetes. The anthropic agency tries to keep the glucose and insulin concentrations of the diabetic patient (who cannot do it by his own) as close as possible to the concentrations of a normal person utilizing a pump that infuses insulin in the patient’s blood. The complete description of the anthropic agency architecture is beyond the scope of this paper. Here, we discuss only the anthropic majordomo agent that manages and supervises the anthropic agency and that provides the unique connection with other external devices. The anthropic majordomo monitors the anthropic agency activity (it periodically checks the physiological values read from the patient and the state of the anthropic agency components) to detect both hardware and biological alarm situations. The anthropic majordomo includes a user interface module that is running on a palmtop computer (HP Jornada 548). This choice has two strong motivations: on the one hand, the palmtop can be easily carried with the patient; on the other hand, it represents a way to increase the participation, the interaction, and the awareness of the patient in the activities of the system. The anthropic majordomo is implemented in java and exploits jini [7] to connect with other agencies, as described later. In the following of this paper, we focus on four scenarios that are related to four problems that the anthropic majordomo can detect and pass to the environmental agency:

– the insulin tank is empty: this is triggered when the insulin level in the pump’s tank is under a safety threshold; – an hardware component is malfunctioning: this is triggered when the state of a component is out of given guard ranges; – the biological state of the patient is alarming: this is triggered when the control activity of anthropic agency is not effective and the glucose level in the patient’s blood cannot be kept enough close to that of a normal subject; – a telemedicine consult is required: this is triggered when the user explicitly requests a telemedicine consult with a medical expert or when the system is addressing the solution of the problems related to a biological alarm (see previous scenario). We selected these four scenarios because they represent important problems for diabetic patients covering both hardware and biological issues, according to the taxonomy outlined in Section 1. The solutions the environmental agency gives to these problems are discussed in Section 4.

3

The Environmental Agency

The environmental agency is built up by agents that are pervasively distributed in an environment. For our purposes, an environment can be a room, an apartment, a building, and so on; moreover, an environment (such as a building) can be composed of several sub-environment (such as the rooms). The architecture of the environmental agency provides solutions that address the volatile and heterogenous composition (the agency is composed of heterogeneous agents that dynamically join and leave; for example when a person bearing monitoring devices walks through a room equipped with environmental devices), the events management (the agency manages reactively the events generated by agents), and the problem solving (the agency solves the exigencies using the functions offered by the agents that are currently connected). The agents of the environmental agency are heterogeneous: they have programming languages, hardware implementations, and communication protocols. In order to cope with this problem we arranged each agent in two parts: the cooperative semiagent (CO) and the operative semiagent (OP). (See [8] for the same idea applied to robotic agents.) The role of CO is to make the network of agents homogeneous. All the COs of the environmental agency have the same architecture. Each CO, implemented in java, provides functions for the connection and the disconnection to and from the agency, for the events management (the agent can generate an event, register itself to particular events, and receive the notification that a registered-to event happened), for the goals management or problem-solving ability (the agent can generate a goal, propose a decomposition for a goal, and carry out an action to solve a goal), and for the communication ability to send messages to the other environmental agents. The OP is the operative and particular component of the agent. The OPs are the pervasive devices scattered in the environment. Their specific functions

can be computational or interactive with the physical world. An OP can be implemented with different architectures and by different programming languages. Thus, it is usually necessary to introduce an additional component commonly employed in multiagent systems, called wrapper, that maps the functions the OP can perform to java methods accessible from the CO (Fig. 1) and that facilitates the reuse of the devices in different applications.

CO State ReadState()

void SetPower(PowerRange Value)

wrapper OP (heating device)

Fig. 1. The wrapper that makes the functions of an heating device available as java methods to the CO, in order to form an heating agent

A particular environmental agent, called environmental majordomo agent, manages and supervises the environmental agency. The environmental majordomo, implemented in java, exploits jini [7] to manage the network of agents. It keeps information about the dynamic registration of the agents to the environmental agency. In fact, the COs of the environmental agents can automatically discover the presence of the environmental majordomo and join the environmental agency by the discovery and join protocols of jini. In addition, the environmental majordomo is the core of both the events management and the problem solving activity of the environmental agency, as explained in the following section. The anthropic majordomo described in the previous section has been completed with a CO for allowing it to be part of the environmental agency. Once a connection between the two majordomos - anthropic and environmental - is established, the anthropic agency is, from the environmental agency’s point of view, an agent. Conversely, from the anthropic agency’s point of view, the environmental agency can be seen an extension of its functions. Such connection is only temporary and can last for few seconds (e.g., when the user is walking through a room) or for several hours (e.g., when the user is at bed in a hospital room). In order to guarantee the correct interpretation of messages in heterogeneous multiagent system, a message must include the ontology [9] adopted to compose the content. In our implementation, we used messages very similar to those of FIPA ACL [10]. The effort to have complete FIPA compliant messages will be little and we will undertake it when the formal semantics of FIPA ACL will be stable. We also devised a simple ontology for our experimental setting, since we

could not find any adequate existing ontology that covers both the medical and the commercial issues we need in our messages.

mechanical and biological exigencies detection

(repository agents) (communication agents)

(interactive agents)

anthropic agency

SMS agent

email agent

fax agent

phone agent

address book agent

OP

OP

OP

OP

OP

OP

OP

OP

OP

CO

CO

CO

CO

CO

CO

CO

CO

CO

store agent

thermometer agent

heating agent

anthropic majordomo

communication network connection, disconnection event management (generate, register, notify) problem solving (generate goals, propose decompositions, execute actions) communication

CO

OP

agent connection management event management problem solving

environmental majordomo agent

Fig. 2. The architecture of the environmental agency we implemented

Our current implementation of the environmental agency is composed of three main classes of agents: communication agents, repository agents, and interactive agents (see Fig. 2). The communication agents provide communication capabilities and include the SMS agent that embeds a toolkit [11] for sending and receiving SMS (Short Message Service) to and from simulated cell phones; this toolkit enables the use of the push technology for SMS to send “clickable” messages to the user, who will have the option of getting more detailed information with a few clicks: in this way a sort of “structured conversation” can be carried out. Other communicative agents are the email agent, fax agent, and phone agent for sending and receiving emails, faxes, and phone calls. The repository agents keep information or physical goods; they are the address book agent, a database of contacts, and the store agent, a database of medicines currently present in the environment. The interactive agents provide to the agency the passive (sensors) and active (actuators) interaction with the environment; they include the thermometer agent, a temperature sensor, and the heating agent that can change the temperature in the environment. The environmental agency we implemented is completed by the environmental majordomo and by the anthropic agency connected via the its majordomo. The current implementation of

the environmental agency simulates all the physical devices involved (cell phones, repositories, thermometers, . . . ). Finally, note that the environmental agency is designed to allow a multilevel composition of agencies. In the same way the anthropic agency is viewed by the environmental agency as a single agent, the environmental agency itself can be viewed (via its environmental majordomo) as a single agent by another environmental agency at an upper level. For example, it is possible to have an environmental agency in every room of an hospital and an environmental agency for every floor of the hospital that views the environmental agencies of the rooms at that floor as single agents.

4

Coordination and Cooperation Mechanisms

In this section we describe the coordination and cooperation mechanisms developed in the environmental agency. Firstly, we illustrate the general concepts of these mechanisms and then we present the details of our experimental activity oriented to glucose-insulin metabolism monitoring. 4.1

Events Management and Problem Solving

In what follows we describe the two basic mechanisms by which the environmental agency coordinates the activities of the agents. The first mechanism manages the events. The environmental majordomo collects the registrations of environmental agents that declare their interest in being informed about an event. When an environmental agent generates an event e, it requests to the environmental majordomo the list L(e) of the listeners for e and notifies e to the agents in L(e). Note that the notification of an event is done by the generating environmental agent and not by the environmental majordomo; this requires one less communication of the data describing the event. These data include the name of the event, the ontology in which the event is interpreted, and a parameter (a java object) that contains all the important information from the event generator to the listeners. As an example of the use of events, consider the case in which the SMS agent receives a SMS that answers a previous SMS it sent. The SMS agent generates the event SMSResponse, requests to the environmental majordomo the list L(SMSResponse) of the listeners for that event, and then sends the event notification to the agents in L(SMSResponse). In this way, an environmental agent that previously requested to send a SMS and subscribed for the SMSResponse event receives the answering SMS. The management of events gives the environmental agency a reactive behavior that allows it to quickly and effectively cope with sudden changes in the environment. The second mechanism is centered around a planner for problem solving. Also this activity is under the supervision of the environmental majordomo. We employ a typical backward planning system of artificial intelligence that operates in the plan space [12], improved to tailor the plans it produces to the agents currently present in the environment. When an agent wants to solve a problem,

it requests the environmental majordomo to reach a goal G that expresses the problem. The problem solving process is articulated in the following steps. 1. The environmental majordomo receives the request about G and asks all the currently connected agents to send the decompositions they know for G. A decomposition brings an initial goal G to a set of subgoals {G1 , G2 , . . . , Gn }. Each one of these subgoals can be either primitive or non-primitive (composite subgoal). In the first case, the subgoal is a simple operation that can be directly executed by the agent proposing the decomposition. In the second case, the subgoal needs further decompositions according to a backward approach. The subgoals generated by a decomposition may be constrained by three different operators. The ordering operator imposes a sequential constraint over the execution of the subgoals. The selection operator imposes that a goal is executed only if a condition holds. Finally, the iteration operator imposes that a goal is executed for a given number of times or until a condition becomes trues. 2. The environmental majordomo collects the possible decompositions for G and picks up the best one. Each decomposition has associated three numeric indexes that express the efficiency, the cost, and the probability of success of the decomposition; such indexes values are goal dependent. For example, if the goal is SetTemperature, the efficiency of a decomposition can be the (estimated) time it requires to reach the target temperature while the cost can be the electric power required to reach the target temperature. If the environmental majordomo does not receive any decomposition for G, it backtracks to select another decomposition among those proposed in the previous steps. 3. The application of the selected decomposition to G generates a tree of goals and subgoals, called planning tree, whose root is G and whose leaves are either primitive operations or new (non-primitive) subgoals that call for further decompositions. When the leaves are all primitive operations, the planning tree is said to be complete. The environmental majordomo iterates the steps 1 − 3 on non-primitive subgoals until the planning tree is complete. The planning tree is expanded with a breadth-first search method. 4. When the planning tree is complete, the environmental majordomo executes the plan by coordinating and supervising the environmental agents that carry out the actions prescribed by the plan. The execution of a plan is a recursive process that can be described as follows: (a) the execution of the initial goal and of a non-primitive subgoal consists in the execution of all the subgoals derived from it, according to their associated operators (ordering, selection, iteration); (b) the execution of a primitive subgoal consists in requesting the execution of the simple action associated with it to the proposing agent and waiting for the output of this execution. The execution of a plan is thus reduced to finding a legal sequence of operations within a complete planning tree. Examples of the problem solving activity of the environmental agency are shown in the next subsection. We remark that the initial goal and the subgoals have

input and output parameters that are propagated to (mapped on the parameters of) the subgoals. For example, the goal SetTemperature has an integer input parameter that specifies the desired target temperature. Goals can have also output parameters, as in the case of the goal SearchCellNumber that has an input string parameter specifying a name or a keyword and a string vector output parameter specifying the mobile phone numbers associated to the name or the keyword. 4.2

Experimental Activity

This subsection presents some of the results obtained from the experimental activity on the environmental agency we implemented. Here we aim to show the effectiveness of the coordination and cooperation mechanisms and not the flexibility in managing the connction of agents. Hence, we suppose that all the environmental agents (including the anthropic majordomo) presented in Fig. 2 are connected to the environmental majordomo. With respect to the four scenarios discussed in Section 2, the planning system tries to maximize the efficiency index (during step 2 of the problem solving process). When in the first scenario, namely when the level of insulin in the tank of the anthropic agency is under a safety alert level (set by the user), the anthropic majordomo detects this situation and generates the goal CheckAndRequest for the environmental agency. The environmental majordomo asks the agents currently connected to the environmental agency to send their available decompositions. The only decomposition for CheckAndRequest is provided by the environmental majordomo itself and brings to two subgoals, as shown in Fig. 3.

CheckAndRequest

IsThere

False

Request

store agent

CreateRequestText SendSMS SearchCellNumber

anthropic majordomo

address book agent

SMS agent

Fig. 3. The complete planning tree for the goal CheckAndRequest

The two subgoals are connected by a selection operator: if the output of the execution of the subgoal IsThere (that checks if some insulin is left in the storeroom) is False then the subgoal Request (that requests pharmacies to supply

insulin) is executed. The planning process at this point starts again trying to decompose the goal IsThere. This subgoal can be decomposed only by a primitive subgoal of the store agent. Then, the subgoal Request is considered. The phone, SMS, fax, and email agents propose their decomposition of the subgoal in order to contact the pharmacies by different communication means. Since the efficiency of the decomposition proposed by the SMS agent is the highest, this is selected. This decomposition constraints the subgoals CreateRequestText and SearchCellNumber to be executed before SendSMS, by an ordering operator. The first subgoal creates the text containing the request to supply insulin and the parameters that the pharmacies have to specify: delivery time, cost, and so on. The second subgoal finds the mobile phone numbers of the pharmacies. The SendSMS subgoal sends out the requests. The planning process continues and the new generated subgoals are decomposed in primitive operations performed by the anthropic majordomo, the address book agent, and the SMS agent. The plan is now complete and ready to be executed. After the execution, if pharmacies have been contacted the anthropic majordomo (that generated the initial goal CheckAndRequest) registers to the SMSResponse event to receive the answers from the pharmacies. When the offers from the pharmacies have been received (or after a timeout), the anthropic majordomo takes its decision about where to buy insulin. This decision is based on the proposed delivery time, on the estimated time left to empty completely the insulin tank, and on the price. The decision is actuated by generating a goal Order that, when expanded and executed, sends a SMS to the selected pharmacy to confirm the order. In the second scenario, the malfunction of an hardware component of the anthropic agency is solved similarly to the empty insulin tank problem. The anthropic majordomo generates the goal Fix that, once decomposed with the heuristic of maximizing the efficiency, sends a SMS to the technical experts of the hardware component not working properly. The experts can send back some suggestions to cope with the situation triggering the generation of new goals by the anthropic majordomo. In the third scenario, when the anthropic majordomo detects that something is wrong in the physiological parameters of the diabetic patient, it generates different goals, according to how critical the situation of the patient is. The criticality depends on the physiological parameters values and on the consciousness test performed by the anthropic majordomo (the palmtop emits a loud sound and waits for the answer of the user). If the situation is not very critical, then a Telemedicine goal is generated (see below); otherwise an emergency message (by SMS, phone, fax, or email) is sent to the closest hospital found in the address book. The last scenario we considered, the telemedicine, requires the most intensive knowledge management. The goal Telemedicine generated by the anthropic majordomo (after a biological alarm or an explicit request by the user) can be decomposed in various ways according to the employed communication mean. Since we maximize efficiency, the sending of a SMS to a medical expert is pre-

ferred. This message contains the values of the main physiological parameters acquired by the anthropic agency. The medical expert can send back suggestions (e.g., to raise the temperature); these are interpreted by the anthropic majordomo that can generate new goals to satisfy them. Moreover, the medical expert can ask for more data. This request is interpreted by the anthropic majordomo that collects and sends to the medical expert the required data (e.g., the insulin and glucose levels during the last day).

5

Related Works

The system we presented in this paper can be thought as an application of the paradigm of the ubiquitous computing [13] to a medical context. In this perspective, the environmental agency can be considered from three different points of view in relation to the current state of the art. These are related to personal area networks (PANs), to smart environments, and to telemedicine. The adoption of PAN [14] in medical applications has been deeply investigated with respect to the cardiological applications. [15] developed a system that acquires signals from the sensors of three leads, processes such signals to obtain (in real-time) the ECG graph, shows the graph on the screen of the palmtop, and then transmits it to a remote physician. [16] designed and implemented a software, called Sutil+, that allows the physicians to receive on their palmtops the electronic health records of the patients. Sutil+ is focused on the automation of messages exchanging in hospitals without any data analysis support. From several other papers (for example, [17] and [18]), it emerges that, in medicine, the palmtops (in PAN) are adopted to record and monitor the health state and the clinic information of the patients and to show such information to both the patients and the physicians. We propose an innovative use of the palmtop as a bridge between anthropic and environmental devices. In part complementary to the concept of PAN is the concept of smart environment. In the medical context, the Smart Medical Home [19] is a particular home that is “aware” of the human activity inside it and that reacts to the perceived state by activating particular devices. [20] proposes a smart home that monitors some biological signals (i.e., ECG and EEG) when a person is in particular places of the home: bed, lavatory, and bathtub (all these places are equipped with smart devices). [21] designed a device that monitors some cardiac functions of the people living in a house. This device receives ECG and blood pressure signals from a wearable computer, processes the signals, recognizes the presence of some specific situations of arrythmia, and when such situations show up the device transmits all the data to a hospital monitor center via the Internet. Our environmental agency provides mechanisms for more complex (e.g., goal oriented) behaviors. We report just the most important general projects that address telemedicine [22]. The MOMEDA system [23] records the data of the patients and sends them to an hospital center. The system for the remote monitoring of the diabetic patients proposed by [24] allows a patient to send his glycemic values to a diabetes

specialist. Finally, we cite the system for managing the emergency calls proposed by [25]; differently from ours that is able to decide autonomously when to make an emergency call, it needs to be activated by a person.

6

Conclusions

In this paper, we have illustrated a cooperative multiagent system for managing the information coming from body implanted devices. We demonstrated by simulation the feasibility and the potential of such system in the case of glucose-insulin regulation in diabetic patients. We implemented in a real device (a palmtop) part of the most important agent, the anthropic majordomo, that plays the role of a “bridge” between the anthropic agency and the environmental agency. Future research will be carry out along two directions. On the one hand, we aim to refine the distributed planner developing new problem solving strategies and to adopt real mobile phones and real devices for interaction with the human body and with the environment. On the other hand, we are trying to further evaluate the possibilities our approach offers in the cardiopathologies context.

References 1. Ross, D.A., Sanford, J.A.: Wearable computer as a remote interface for people with disabilities. In: Proceedings of the First IEEE International Symposium on Wearable Computing, Boston, MA, USA (1997) 161–162 2. Amigoni, F., Somalvico, M., Zanisi, D.: A theoretical framework for the conception of agency. International Journal of Intelligent Systems 14 (1999) 449–474 3. Amigoni, F., Dini, M., Gatti, N., Somalvico, M.: Anthropic agency: a multiagent system for physiological processes. Artificial Intelligence in Medicine 27 (2003) 305–334 4. Mocomed. http://www.mocomed.org/ (2002) Last accessed September 3rd, 2002. 5. Wooldridge, M.: An Introduction to Multiagent Systems. John Wiley and Sons, Chichester, UK (2002) 6. Cobelli, C., Nucci, G., Prato, S.D.: A physiological simulation model of the glucoseinsulin system. In: Proceedings of the First Joint Meeting BMES/EMBS Conference. IEEE Computer Society Press, Atlanta, GA (1999) 7. JINI. http://wwws.sun.com/software/jini/ (2002) Last accessed July 9th, 2003. 8. Amigoni, F., Somalvico, M.: Dynamic agencies and multi-robot systems. In Lueth, T., Dillmann, R., Dario, P., Worn, H., eds.: Distributed Autonomous Robotic Systems 3. Springer-Verlag (1998) 215–224 9. ONTOLINGUA. http://www.ksl.stanford.edu/software/ontolingua/ (2002) Last accessed July 9th, 2003. 10. FIPA Communicative Act Library Specification. http://www.fipa.org/ (2002) Last accessed July 9th, 2003. 11. Nokia Mobile Internet Toolkit. http://www.nokia.com/ (2002) Last accessed July 9th, 2003. 12. Nilsson, N.: Artificial Intelligence: A New Synthesis. Morgan Kaufmann, San Francisco, CA, USA (1998)

13. Weiser, M.: Some computer science issues in ubiquitous computing. Communications of the ACM 36 (1993) 74–84 14. Jovanov, E., Price, J., Raskovic, D., Kavi, K., Martin, T., Adhami, R.: Wireless personal area networks in telemedical environment. In: Proceedings of the IEEE Fourth International Symposium on Wearable Computing, Atlanta, GA, USA (2000) 22–27 15. Zou, Y., Guo, Z.: A palm pilot based pocket ECG recorder. In: Proceedings of the IEEE EMBS International Conference on Information Technology Applications in Biomedicine, Arlington, VA, USA (2000) 110–112 16. Prescedo, J., Castro, D., Vila, J., Fern´ andez-Delgado, M., Fraga, S., Lama, M., Barro, S.: Wireless interface for monitored patients in coronary care units. In: Proceedings of the IEEE 22nd Annual International Conference on Engineering in Medicine and Biology Society, Chicago, USA (2000) 17. Martin, T., Jovanov, E., Raskovic, D.: Issues in wearable computing for medical monitoring applications: a case study of a wearable ECG monitoring device. In: Proceedings of the IEEE Fourth International Symposium on Wearable Computing, Atlanta, GA, USA (2000) 43–49 18. Franchi, D., Belardinelli, A., Palagi, G., Ripoli, A., Bedini, R.: New telemedicine approach to the dynamic ECG and other biological signals ambulatory monitoring. In: IEEE Computers in Cardiology. Volume 25., Cleveland, OH, USA (1998) 213– 216 19. Center for Future Health. http://www.centerforfuturehealth.org/ (2002) Last accessed July 9th, 2003. 20. Kawarada, A., Takagi, T., Tsukada, A., Sasaki, K., Ishijima, M., Tamura, T., Togawa, T., Yamakoshi, K.: Evaluation of automated health monitoring system at the “welfare techno house”. In: Proceedings of the IEEE 20th Annual International Conference on Engineering in Medicine and Biology Society. Volume 4., Honk Kong (1998) 1984–1987 21. Bai, J., Zang, Y., Sheng, D., Wen, L., Ding, C., Cui, Z., Tian, F., Yu, B., Bing Dai, J.Z.: A portable ECG and blood pressure telemonitoring system, design and development aspects of a cost-effective system for home-care telemedicine applications. IEEE Engineering in Medicine and Biology Magazine 18 (1999) 63–70 22. Reponen, J., Niinimaki, J., Holopainen, A., Jartti, P., Ikko, E., Karttunen, A., Kumpulainen, T., Tervonen, O., Paakko, E.: Momeda, a mobile smartphone terminal for dicom images and web based electronic patient data. Journal of telemedicine and telecare 6 (2000) 45–49 23. MOMEDA. http://www.biomed.ntua.gr/momeda/ (2002) Last accessed July 9th, 2003. 24. Pattichis, C.S., Kyriacou, E., Voskarides, S., Pattichis, M.S., Istepanian, R., Schizas, C.N.: Wireless telemedicine systems, an overview. IEEE Antenna’s and Propagation Magazine 44 (2002) 143–153 25. Antoniades, C., Kouppis, A., Pavlopoulos, S., Kyriakou, E., Kyprianou, A., Andreou, A.S., Pittichis, C., Schizas, C.: A novel telemedicine system for the handling of emergency cases. In: Proceedings of the 5th World Conference on Injury Prevention and Control, New Delhi, India, World Health Organization (2000)