Fish disease diagnosis program — problems and some solutions

Abstract

Any software dealing with disease diagnosis has to overcome various problems. Some are inherent in the diagnostic technique, others arise because of the specific problem domain. We have evaluated different expert-system technologies including neural-nets, case-based expert systems (ES), rule-based ES and fuzzy logic. The problem domain (fish disease) has it’s own problems, the major one being that there is no accepted database of cases like there is in other medical fields. This precludes the use of diagnostic techniques needing a large number of test cases. The other problem in this context is the effort to deal with ALL diseases for multiple species. We explore the different ES techniques, and outline the final product (Fish-Vet) which includes a hybrid system that enables us to obtain reasonable diagnoses in a timely manner. This program uses elements of fuzzy, rule-based and statistical systems. The mix and match approach proved useful, and further work has to be performed in order to incorporate other artificial intelligence techniques into the process.