[Differential diagnosis using artificial neuronal networks]

In order to rule out the effects of subjective factors and decrease the number of diagnostic errors, artificial neuron nets are proposed. By means of these nets the subjective and half-empirical heuristics are replaced with rational diagnostic information based on quantitative and logic analysis. The proposed method was used for deriving the decisive regularities ensuring the differential diagnosis between infective endocarditis and active rheumatic fever, infective endocarditis and systemic lupus erythematosus, and systemic lupus erythematosus and active rheumatic fever. Reliability of diagnostic tables representing the decisive regularities in the usual form is confirmed by clinical data.