Biomedical applications of uncertainty modeling and analysis with fuzzy receiver operating characteristic methodology

Receiver Operating Characteristic (ROC) methodology is popular and well-established in medicine and other disciplines. It applies where a variable is used for two-state classifications. Fuzzy ROC methodology introduced earlier by the authors as an extension of ROC methodology for fuzzy classifications, is further developed in this paper. Of particular interest here is the idea of modeling an individual´s fuzzy membership with one or many, numerical or ordinal values, the use of the bootstrap to obtain nonparametric confidence intervals for all ROC plot parameters, the notion of using ROC-derived fuzzy functions to estimate degree of membership when a gold standard is unavailable, and the use of fuzzy ROC methodology to enhance artificial neural network performance. Medical examples with fever, macrosomic newborns, and adverse drug reactions as fuzzy membership classes are used to demonstrate basic concepts