Transfer approach to the training of convolutional artificial neural networks in the task of diagnosing pulmonological diseases

Maslennikov Vladimir Vladimirovich, Daeva Sophia Georgievna

The article is devoted to the study of the application of the transfer approach to the training of convolutional artificial neural networks when working with medical graphic materials. An algorithm for developing a classifier based on convolutional neural networks for detecting pneumonia on digital images of chest X-rays is proposed. The study shows the advantages of the considered approach to the training of convolutional neural networks, as well as the advantages of the classifier developed on its basis over analogues.

Keywords:artificial intelligence, transfer learning, convolutional neural networks, pneumonia detection, classifier.