The research study of the adaptive neuro-fuzzy interference system (ANFIS) for the diagnostics of endogenous intoxication syndrome with chronic kidney disease

Abstract

The article presents information on the application of the neuro-fuzzy diagnostics method of endogenous intoxication syndrome (EIS) in patients suffering from the chronic renal failure and undergoing long-term ambulatory hemodialysis. Early diagnostic task of the ndogenous intoxication syndrome is of great importance, as the prevalence rate of chronic kidney disease (CKD) is not less than 10%, reaching 20% or more in certain categories of persons.The publication offers a less expensive and more accessible diagnostic method of EIS. The method is based on the application of the neuro-fuzzy network. The systems based on neuro-fuzzy networks make conclusions on the basis of the knowledge data base which contains a priori expert’s experience, and the membership functions parameters are configured using learning algorithms of neural networks.They used an adaptive neuro-fuzzy interference system (ANFIS), implemented in the software environment “Matlab”. The article presents the structure of fuzzy neural network consisting of five layers for the diagnostics of EIS. Specified parameters and actions carried out on each network layer are described in details. A fuzzy knowledge data base is set up for modeling of the patient’s state. The knowledge data base includes 18 rules of fuzzy production that correspond to the consistency condition. For ANFIS network learning a back propagation of error algorithm was used, which allows to prepare better the neuro-fuzzy network for solving of specific problems in less time.Conducted experiments showed that by virtue of the use of the neuro-fuzzy network for the diagnostics of endogenous intoxication syndrome, it became possible to provide a sufficiently high accuracy of the diagnostics.