DEVELOPMENT OF NEURAL NETWORK CLASSIFICATOR OF DISEASES UROFLOWMETROGRAMS IN UROLOGY

N.I. Fedorenko

The neural network method of detecting uroflowmetrograms in urology is suggested. Basing upon this method, by using neural networks of counterpropagation, the neural network uroflowmetrogram classificatory is developed. The approbation of such approach was conducted for four types of illness. By solving this problem, possible variants of using alternative architectures of neural networks were analyzed. Neural network learning was conducted with 120 input vectors basing upon real data that characterize typical urologic illness. Testing of the network was carried out with 30 different input vectors.

Keywords: uroflowmetrograms, types of illnesses in urology, architectures of neural networks, neural network illness detection, counterpropagation networks, neural network uroflowmetrogram classificator.