Agent Technology and Healthcare: Possibilities, Challenges and Examples of Application - Computational Intelligence in Healthcare: Advanced Methodologies - page 37

owns all the information related to the services in which it is involved and manages
all queries and requests coming from the associated user or other Actor Agents. They
also incorporate all the details about user's roles and permissions according to the
organizational data available in the knowledge bases (specifically, the Actor Profile
Ontology). In fact, Actor Agents ' code is partially created in an automated fashion
from the knowledge stored in those structures. Concretely, agent behaviour skeletons
are created from the skill definitions stored in the ontologies. In addition to this
“static” knowledge, it is also possible to adapt this data in order to personalize the
interaction between the user and the system at runtime. This tailoring process is medi-
ated by his/her associated Actor Agent in function of the locally stored profile-related
and organizational knowledge.
In order to provide access transparency to end users, the Web interface and the
multi-agent system are connected by a bridge constituted by a servlet and a Gateway
Agent (GA). When an actor logs into the system, the servlet creates a Gateway Agent
whose mission is to keep a one-to-one connectivity with the corresponding permanent
agent. Considering that the Web and agent environments are not meant to be directly
interoperable, GAs act as mediators, translating web request to FIPA-based agent
messages. This provides independency between the ubiquitous web accesses from the
user's PC or mobile device and the Actor Agent's implementation and execution.
Ideally, Actor Agents should be executing in a secure environment and distributed
through a computer network in order to provide load balancing in conditions of high
concurrency.
Finally, SDA* Executor Agents (SDA-Es) allow enacting a care plan by recom-
mending the next step to follow according to the patient's current state. The SDA-E is
dynamically created by an Actor Agent in order to enact a SDA* structure correspond-
ing to a healthcare process (like an Individual Intervention Plan) or to a management
procedure. The SDA-E agent, after loading a SDA* from the repository, is ready to
receive queries about its execution. Those queries are performed by the Actor Agent in
order to know the next step to follow. As a result, healthcare actions will be delegated
to the appropriate actors considering the organizational knowledge stored in the on-
tologies and the availability of physical actors represented by means of Actor Agents.
Clinical decisions which may arise during the execution of an SDA* will be forwarded
to the user in order to implement the healthcare treatment in a supervised manner.
In the same manner as HeCaSe2, the agent-based platform has been implemented
with the JADE tool, and using two FIPA-based protocols: FIPA-Request to request
actions to be performed to the SDA-E Agent and FIPA-ContractNet to negotiate the
execution of a particular action between the available partners.
5.3 Discussion
One of the most interesting aspects of the K4Care platform is the decoupling of the
system's implementation from the knowledge structures which represent the HC
model and its particular representation on a concrete organization. In this way, the
medical and organizational knowledge can be directly modified in order to introduce
changes in the HC process, which will immediately be adopted by the actors of the
system in execution time. This design provides a high degree of flexibility, reusability
and generality, which are fundamental aspects of a system modelling complex and
dynamic processes. As an example, in K4Care, the procedures used in potential