A two-dimensional fuzzy repertory grid approach for building medical expert systems

Only single fuzzy relationship between symptoms and diseases is considered in conventional fuzzy repertory grid approach (FRG) to developing medical expert systems. In the preliminary study, a two-dimensional fuzzy repertory grid approach (TDFRG) is proposed to represent the dual-fuzzy relationships between symptoms and diseases for hepatitis diagnosis. One is the severity degree of symptoms; the other is the importance degree of symptoms. Moreover, both patient subjective indications and blood test items are considered as symptoms in this paper. To evaluate the validity of the proposed TDFRG approach, and compare the performance of different feature subsets for hepatitis diagnosis, six experiments are conducted and the performance comparisons of the proposed TDFRG approach with conventional FRG method are reported.