Pre-trained Convolutional Neural Networks for the Lung Sounds Classification

Abstract

This article discusses the use of pre-trained sophisticated architectures of convolutional neural networks for the classification of lung sounds. The classification was made on the basis of the spectrogram of lung sound signals. The images of the spectrograms were analyzed directly, a set of parameters was obtained, on the basis of which the neural networks were trained. As the results show, the proposed classification method can serve as a good auxiliary tool for making a diagnosis to help the pulmonologist.