I. INTRODUCTION
Image segmentation is the process of delineating structures of importance from an image. Identifying these structures in the medical image finds application in many medical procedures. To state some of them: 1) segmentation of optic cup and disc in the retinal fundus image is useful in glaucoma screening, 2) segmentation of polyp in colonoscopy image is helpful in cancer diagnosis, 3) segmentation of the organ, bones benefit surgery planning and 4) segmentation of lung nodules in chest Computed Tomography aids physicians to differentiate malignant lesions from benign lesions. In recent years, deep learning networks [1] are widely used in medical image segmentation, and the most commonly used deep learning network is UNet [2].