Broker Architecture for Intelligent Agent Subscription in ICU Sofie Van Hoecke, Kristof Steurbaut Supervisor(s): Filip De Turck, Johan Decruyenaere, Bart Dhoedt Abstract— Web services are becoming the standard for integrating heterogeneous software components. In order to support dynamic selection and composition of Web services, based on criteria such as minimization of response time and maximization of fault tolerance, there is a strong need for a service broker architecture. Since the Intensive Care Unit (ICU) is an example of an extremely data-intensive environment, an ICU broker architecture can facilitate the abstraction of relevant information and support the physician through software agents for medical decision support. It is expected that in future ICU systems, tens of agents will be active simultaneously in order to optimize the care of critical ill patients. Therefore the authors present a broker platform offering advanced features such as transparent data migration; user-friendly patient/agent subscription and profile based filtering of support messages. A platform prototype is currently being evaluated by the Department of Intensive Care of the Ghent University Hospital. Keywords— Web services, broker architecture, medical support agents, ICU.
I. I NTRODUCTION An ever increasing number of applications use the Internet to interact with each other without human intervention. By exchanging XML-messages, complying with the Web service standards, previously incompatible platforms can easily integrate. Many times in the past, this goal of perfect interoperability has been promised, although it has been rarely achieved. Since many major vendors are pushing Web services and have agreed to common XML-based standards, Web services are a successful concept for the integration of heterogeneous software components. Based on the exchange of structured text messages, the interaction abstracts the underlying technologies.
is reduced to authenticating (e.g. by means of eID, smart card or login/password) and selecting one of the (composed) services. The presented architecture covers a wide range of application cases. For example multimedia content delivery [1] can benefit from QoS brokering. The broker middleware can dynamically select and compose the needed services (e.g. services for broadcasting, streaming, payment and security) in order to set up a video-on-demand stream meeting the request (e.g. high quality, no delay or limited output device). Another case can be found in eCommerce, where a call center for example negotiates with multiple credit checkers, in order to acquire payment validation. Based on the call center load, the broker middleware can divide the requests over multiple credit checkers in order not to lose or displease clients. QoS brokering can also be applied in B2B (Business to Business) for optimizing the virtual supply chain of delivery companies. QoS brokering can select delivery services with the best QoS (e.g. delivery time, quality, price), resulting in smaller stocks and advanced efficiency. On the other hand, QoS brokering can also be useful in pure software design, planning and measuring the required infrastructure, as well as in offline tuning of load balancing and brokering strategies. Finally eHealth is another case that can benefit from QoS brokering by integrating multiple care providers [2] [3]. The deployment of the QoS brokering middleware for eHealth, more in particular for ICU, is described in this section.
II. N EED FOR W EB S ERVICE B ROKER
III. B ROKER P LATFORM FOR I NTELLIGENT AGENT S UBSCRIPTION IN ICU
As already stated, extending existing services with a Web service interface enables integration. Web service composition, resulting in the deployment of multiple promising end-to-end and eBusiness applications, is enabled through the emerging Web service standards on the one hand for flow specification, such as the Business Process Execution Language for Web Services (BPEL4WS), and on the other hand for semantic markup, such as the Web Ontology Language for Web services (OWL-S). However QoS (Quality of Service) requirements such as response time, availability and price can influence and hamper service composition. Therefore we present a middleware platform for brokering of composed services with QoS guarantees. Implementing this brokering middleware by means of Web service technology creates the required integration of heterogeneous services, taking into account QoS requirements, and offers an advanced set of composed services crossing multiple service and data providers. In this way the required user interaction
The Intensive Care Unit (ICU) is an extremely data-intensive environment. This large amount of data exceeds however human intellectual capabilities [4]. Information technology can facilitate the abstraction of relevant information and support the physician through software agents for medical decision support. Typical examples of medical decision support agents for the ICU are a sepsis detection agent and multi-agent systems for prescribing antibiotics. It is however expected that in future ICU information systems, tens of agents will be active simultaneously in order to optimize the care of critical ill patients. Therefore the Web service broker can be applied as a management platform for agent subscription, data handling, the notification and presentation of results from the agents, as well as load balancing of the agents over multiple workstations to execute them simultaneously. The platform is referred to as the Intensive Care Agent Platform (ICAP).
S. Van Hoecke is with the Department of Information Technology, Ghent University (UGent), Gent, Belgium. E-mail:
[email protected].
The ICAP platform has been implemented based on the principles of service-oriented architectures, using the Web service
Suppose a physician wants to be notified immediately if there are changes in the kidney function of his patient. Therefore the physician has to run the client application and authenticate himself. Afterwards he can subscribe to the preferred agent/patient combinations. From then on, whenever a new creatinine value is available from the laboratories, or a new urine measurement is entered in the intensive care patient database, ICAP locates the appropriate agent type for processing these data and selects the best or least loaded server running such an agent. The selected kidney agent is then activated and calculates the new kidney status. If there are any changes in the status compared to the previous one, an alarm is sent to both the bedside monitor of the patient and the physician on his PDA. ICAP facilitates the abstraction of relevant information and support the physician in medical decision making. Due to the improved reaction times, care can be provided faster and at lower cost. IV. C ONCLUSIONS
Fig. 1. ICAP brokering platform. The following components are shown: Trigger Forwarder (TF), Communication Component (CC), Subscription Component (SC), Event Handler (EH), Database Manager (DBM), Agent Manager Coordinator (AMC), Agent Manager (AM), Application Server (AS), Pool Management Server Coordinator (PMSC), Pool Management Server (PMS) and Load Monitor (LM).
technology. The total framework is composed of multiple services as independent building blocks. These services are autonomous, independent of other services and only responsible for their own functionality. The framework acts as a generic communication system in which agents can easily be plugged. The agent functionality can cover other areas than medical decision making. Figure 1 depicts the main software components of the platform. A detailed description of each individual component is given in [5]. The platform supports as well push as pull models. Laboratories or monitors can send new medical data results by triggers using the push model. On the other hand retrieving medical data from databases is supported by the pull system. Processing these data from both models, agents can generate a result set or medical suggestion and send it to the appropriate users over the platform. This way, users are notified by ICAP whenever new medical data is available or alarms are generated. The support and alarm messages can be received bedside, at the patient terminal, or on the PDA of the physician.
It is generally expected that within the near future, Intensive Care Unit computerization including advanced real time and bed-side decision making capabilities, will become essential to guarantee the highest quality of care for every ICU patient. In this paper, we motivated the use of a Web service broker allowing for the intelligent subscription of the medical decision support data and offering advanced features such as transparent data migration; user-friendly patient/agent subscription and profile based filtering of support messages. The architecture has been implemented and is currently tested by the Intensive Care Department of Ghent University. By making use of the Web service technology, generic definition of the interface and the exchanged messages, the platform is largely generic, i.e. as independent as possible of (i) the implementation language of the agents, (ii) the operating system and hardware of the workstations and (iii) the particular type of the agents. Single point-offailures are avoided in the platform in order to ensure reliability. Finally, the platform allows for easy integration and supports priority. ACKNOWLEDGMENTS Sofie Van Hoecke would like to thank the IWT (Institute for the Promotion of Innovation through Science and Technology in Flanders) for financial support through her Ph.D. grant. R EFERENCES [1] S. Van Hoecke, W. Haerick, G. De Jans, F. De Turck, E. Laermans, B. Dhoedt, P. Demeester, Design and Implementation of a Secure Media Content Delivery Broker Architecture, The 2005 International Symposium on Web Services and Applications, Las Vegas, Nevada, 2005. [2] S. Van Hoecke, K. Vlaeminck, F. De Turck, B. Dhoedt, Open Web Services-based Middleware for Brokering of Composed eHomeCare Services, 2005 Middleware for Web Services (MWS’05) Workshop, Enschede, The Netherlands, 2005. [3] S. Van Hoecke, K. Taveirne, K. De Proft, F. De Turck, J. Decruyenaere, B. Dhoedt, An Agent-based Platform for Efficient Telemonitoring Data Processing, Invited Paper, Health Pervasive Systems Workshop (HPS’06), Lyon, France, 2006. [4] Morris AH, Gardner RM. Computer applications. In: Hall J, Schmidt G, Wood L, eds. Principles of critical care. New York: McGraw-Hill, 1992: 500-514. [5] S. Van Hoecke, F. De Turck, K. De Proft, K. Taveirne, C. Danneels, J. Decruyenaere, Platform for Intelligent Agent Subscription in ICU, MedNET, Toronto, October 2006.
