Mysnyk A. One-dimensional and two-dimensional diagnostic data analysis using the methods of artificial neural networks

New artificial neural network model with adding neurons and expert diagnostics method for biophysics analysis and pathology recognition based on such neural net is presented. It is shown, that local minima problem could be solved using additional inter-neuron links along with gradient initialization and entropy optimization method. Expert diagnostics method which is based on proposed method of data analysis is provided and dependencies between textural characteristics of medical images and some types of diseases is investigated. It is experimentally shown, that during pathology state analysis, normal diagnoses are recognized better than deviation and recognition probability does not depend from boundary condition for ultrasound, magnetic resonance and spectra data. Efficiency estimations for developed method are presented and new neural-based medical images segmentation algorithm is proposed. Area of application - medical diagnostics and biophysical systems modeling.