A fuzzy hierarchical approach to medical diagnosis

In this paper a fuzzy hierarchical approach to medical diagnosis is presented. It differs from other fuzzy mechanical diagnosis approaches by at least two important characteristics. First, it operates having as output universe of discourse the possibility degree of a certain diagnosis and not the diagnosis itself and, second, it offers some specific methods of dealing with a large variety of uncertainties involved by a diagnosis. The system we propose is based on diagnosis cells (elementary cells) that depending upon the type of knowledge, can be fuzzy inference systems, neural networks, neuro-fuzzy networks, other type of hybrid systems or even simple fuzzy or crisp mathematical formulas. Applications based on this approach revealed good diagnosis performances.