Аннотирование объектов на медицинских изображениях рентгенографии грудной клетки с применением нейронных сетей

Abstract

Misinterpretation of X-ray images can lead to a worsening of the patient's condition. The purpose of this research was to develop an algorithm for automatic annotation of diseases on the X-ray image in order to improve the accuracy of the analysis of medical images. This paper considers various methods for solving the classification problem. During the research a database of annotated medical images of radiography was compiled. On this basis, a compactly connected convolutional neural network was trained and tested. The classification accuracy of developed algorithm is above 66% for 14 classes of diseases.