Art der Arbeit:

• Masterarbeit Informatik
• Diplomarbeit Informatik

Status:

Themenangebot

Ansprechpartner:

• Prof. Dr. Rainer Unland

Kurzfassung

At the University Duisburg-Essen, a holonic medical diagnostic system based on multi-agents technology and swarm intelligence is developed through the last years.

The system combines the advantages of the holonic paradigm, multi-agent technologies and swarm intelligence in building a highly adaptive, scalable, flexible, reliable and robust Internet-based medical diagnostic system.

The medical diagnostic system is compound of three type of agents: a mediator agent, many diagnosis specialist agents and disease representative agents.

The mediator agent represents the diagnostic system interface to the diagnosis requester and to the physician/specialist evaluator. Several diagnosis specialist agents are becoming part of the ad-hoc holonic structure in order to provide a suitable diagnostic. The disease representative agents are related to each defined disease that can be a diagnosis result. A disease representative agent determines if a set of symptoms matches its own symptom pattern.

In diagnosis retrieval process, the agents follow a specific swarm intelligence method to build a holonic structure that provides one or more possible diagnostics. Each diagnostic returned by the system has a probability to match the true diagnostic that, in common way, is determined by the physician. The mediator agent has the role to encapsulate the symptoms in a computer readable patient pattern that can be processed by holonic system. Also, mediator agent has to display the results of a diagnosis finding process together with the appearance probability of each possible diagnostic.

This thesis has to contribute to the specification of mediator agent application. In a first step, the concept and architecture of mediator agent has to be realized. Afterwards, for development of the system, the design and implementation have to be realized.

The following milestones have to be reached by the thesis:

1. Realization of a Graphical User Interface for symptoms and computer readable patient pattern handling, for diagnostic results display and for evaluation done by a physician;

2. Standardization of diagnosis data necessary within diagnostic finding process, especially during hypothesis formation phase

3. Design and Implementation of an Interface for communication with disease specialist agents of appropriate level;

4. Development of logging data module for all results, evaluated and unevaluated by a specialist;

For implementation of the prototype that has to prove the chosen concept for mediator agent, one of the most common object oriented programming languages, Java and C++ have to be used.