Fuzzy systems approach to diagnosis in the postpartum cow

A computer-based fuzzy systems approach to diagnosis was compared with a standard rule-based approach for ability to diagnose 1 or more of 14 diseases in postpartum cows. A comparison between the 2 systems, using the same 12 input variables (clinical signs and laboratory measurements), resulted in 93% diagnostic agreement. Much of the study compared a fuzzy diagnostic system that had 12 input variables with a standard system that had 17 input variables. The results indicated that the fuzzy diagnostic system was more efficient to design (fewer rules), that the fuzzy system often provided a more accurate diagnosis, and that selection of input variables is critically important to diagnostic accuracy for any computer-based diagnostic system. The apparent advantages of fuzzy logic-based decision systems over standard rule-based methods of modeling the medical diagnostic process warrant continued study to determine limitations and other possible applications.