I. Introduction
Medical image segmentation is an essential step in a wide range of clinical applications. For instance, prostate zonal segmentation is beneficial for treatment planning [1], [2], and polyp segmentation in colonoscopy images can provide valuable boundary information for the further surgery [3], [4]. Recently, manual annotation is common in clinical practice, however, it is labor-intensive and prone to inter and intra-observer variability. Hence, there is a high demand on accurate and reliable automatic segmentation methods to derive quantitative assessment for clinical applications.