Using convolutional neural networks for the analysis of nonstationary signals on the problem diagnostics vision pathologies

Abstract

The paper considers the use of convolutional neural networks for the analysis of non-stationary signals. Electroretinograms (ERG) are used to diagnose complex pathologies of vision. A method is proposed based on data clustering, that allows to extract knowledge from biophysical research data in situations where it is impossible to make an unambiguous diagnosis on the basis of available data, or there are several diseases simultaneously.