Measurement of Cardiothoracic Ratio and Detection of Cardiomegaly in X-Ray Images Using Deep Learning

Abstract

In this study, the cardiothoracic ratio is automatically measured by extracting lung and heart regions in a chest X-ray image and measuring their widths. The proposed method uses a deep learning model based on U-Net++ with VGG19_bn encoders. The results of cardiothoracic enlargement detection using the cardiothoracic ratio measured by the proposed method showed a high degree of agreement with the judgment made by a physician. As a result, the automatic cardiothoracic ratio measurement system using the proposed method contributes to a significant reduction in the time and labor of physicians.