The problem of decision making in diagnoses of oral mucosa lesions is considered. It is shown that computer-assisted diagnosis systems can be very helpful for clinicians in this branch. In the case of the presented problem for oral mucosa lesions an expert knowledge is used in formalizing the initial information, as well as in the estimation of indicants of the patient. The presented method is used as an alternative to avoid many problems associated with probabilistic and statistical approaches. It handles the expert information in a natural way. The diagnosis is made considering the weight of each symptom. It is shown how the proposed method can be explained from the theory of fuzzy relations point of view.