Fuzzy inference for the medical diagnositic logic(Fuzzy Logics in Biomedical Systems)

Abstract

Medical data provides us the valuable information that is useful for medical diagnosis. Medical information is based on the results of blood tests, physical examinations, interviews and medical images such as X-ray images and ultrasonic images. It is often unclear, however, which facets or items of medical information are most valuable, how much weight should be given to each item, and how diagnostic logic can be used with these items. Moreover, in the process of scaling the degree of these items, fuzzy characteristics are encountered. For example, in considering the item of "Shape" in images, it can often be difficult to judge the degree of disorder clearly. The criteria for evaluation of each item differ from doctor to doctor; i. e., judgement becomes subjective. Fuzziness also exists in the diagnostic logic because its framework cannot be clearly defined. One medical doctor will select some items from medical information for use in diagnosis, while another will choose others. Moreover, the weights of these items do not always remain constant. To construct a diagnostic system using images, we applied a fuzzy scale, by which the degree of variation in items is evaluated, and applied fuzzy inference using membership functions of the weights of the items to the diagnostic system. In this paper, a fuzzy inference and defuzzination system for the medical diagnosis are discussed. We propose a few methods of defuzzination having two dimensional evluation which are different from the type of central points of gravity on thye fuzzy subsets for all rules.