Features of image preprocessing and segmentation in the task detection of COVID‐19 by x-rays

Dumaev Rinat, Kiryakov Ivan, Molodyakov Sergey

Pulmonary turbidity is an inflammation in the lungs caused by many respiratory diseases, including the novel coronavirus disease COVID-19. A chest x-ray with such opacities makes areas of the lung invisible, making it difficult to automatically analyze images on them. This paper discusses the methods of preprocessing and segmentation of the lungs in the framework of the task of detecting COVID-19 and other diseases from x-rays. The CNN network for segmentation of the lung region with extreme deviations, methods of data processing and augmentation are presented.

Keywords:lung segmentation, COVID-19, convolutional neural networks, machine learning, diagnosis of pneumonia, image recognition.