Towards Recognition of Pleural Effusion Images

Abstract

early diagnosis of patients' diseases allows to prescribe effective treatment in a timely manner. This article presents the results of research related to the recognition of images of pleural effusions. The purpose of the research is the recognition of images characteristic of pathologies associated with oncological diseases. When recognizing images, convolutional neural networks were used. When developing software, the authors used the TenzorFlow and OpenCV libraries. Image recognition accuracy is 95%. The studies are incomplete; the authors are trying to improve the results of research by replenishing the training sample with new copies of images of pleural effusions and using combinations of pattern recognition methods.