Basic Study on Image Segmentation of the Heart Regions by Semi-Teacher Learning

The purpose of this study is to facilitate the diagnosis of cardiovascular diseases. We perform semantic segmentation of the heart from CT images using Fully Convolutional Neural Networks. In the case of medical images it is difficult to create or obtain label data. Therefore, we studied semi - supervised Learning method combining Adversarial Network and Adversarial Training in order to create highly accurate segmentation model with less label data. We showed that a combination of Adversarial Network and Adversarial Training can create a more accurate cardiac segmentation model even when the number of label data is small.