A multi-agent prototype system for medical diagnosis

Coordination and negotiation among agents are necessary when multiple agents are motivated to make a diagnosis for a patient together. In this paper a model of a multi-agent diagnosis helping system (MADHS) is given, where several knowledge-based systems are considered as cooperative agents in medical diagnoses. Fuzziness and uncertainty have been incorporated into decision trees to form the reasoning mechanism of agents. A novel coordination mechanism is then described, which is able to reach the final diagnosis compatible with both patient's anamnesis and existing medical principles. The model and reasoning mechanisms are implemented using Java, Java agent development framework (JADE), Java expert system shell (JESS) and NRC FuzzyJ Toolkit, and is tested by both traditional Chinese and western medical diagnosis examples. It is anticipated that the proposed system and technologies will be widely used in applicative areas, such as multi-agent medical diagnosis, medical helping, and other automatic diagnosis and decision making systems.