Abstract
We have applied the diagnosis method proposed by D.Norris, B.W.Pilsworth and J.F.Baldwin to the diagnosis of valvular heart diseases. This method uses concepts from fuzzy set theory and consists of two independent methods: discrimination analysis and connectivity analysis. We performed the experiments in order to evaluate effectiveness of the proposed diagnosis methods. Also, we extended the original method to handle partial manifestation of symptoms and severity of diseases by using fuzzy sets. In Addition, we introduced the concept of prototypicalness of patients with a particular disease to improve the performance of the diagnosis. The results of the experiments are very promising for the discrimination analysis. In the best case, we achieved a rate of true positive diagnosis of 81% while maintaining a rate of false positive diagnosis at the low level of 10%. The connectivity diagnosis method proposed by D.Norris, et al. produced disappointing results in our experiments. In exploring the reasons for this, we designed a new algorithm which is both more intuitive and more effective. We report the quantitative results of the experiments.
