Expert systems in uroflowmetry data evaluation

Abstract

Introduction. In the practice of an urologist, it is customary to assess the type of urination by two parameters: most often it is the effective volume of the bladder (V) and the maximum volume rate of urination (Qmax). Since the expert assessment of the digital characteristics of urine flow is often ambiguous, they are not taken into consideration by some doctors and often remain without due attention. Today there is a tendency in medicine to objectify by quantification of clinical parameters. The main technology used to solve the tasks of data processing and analysis, as well as their classification and forecasting, are artificial neural networks. The aim of the work was to develop an expert system of urine flow rate data recognition based on neural network classifier.

Materials and methods. The training of an artificial three-layer neural network of direct distribution occurred according 210 uroflowgrams and a multidimensional vector, characterized by 9 input parameters.

Results. The system was tested on 40 examples ‒ uroflowgram data of patients who did not participate in neural network training. Despite this fact, the neural network has identified all the proposed examples correctly.

Conclusions. A neural network method for recognition of uroflowmetry data of various diseases of the lower urinary tract is proposed. The space of informative features influencing the assessment of uroflowmetry data is formed. An expert system that classifies diseases (3 types of disorders) of the lower urinary tract with a 95% degree of confidence has been developed.