Synthetic Augmentation pix2pix using Tri-category Label with Edge structure for Accurate Segmentation architectures

In medical image diagnosis, pathology image analysis using semantic segmentation becomes important for efficient screening as a field of digital pathology. The spatial augmentation is ordinarily used for semantic segmentation. Images of malignant tumor are rare, and annotating the labels of the nuclei region is a time-consuming process. An effective use of the data set is required to maximize the segmentation accuracy. It is expected that augmentation to transform generalized images influences the segmentation performance. We propose a synthetic augmentation using label-to-image translation, mapping from a semantic label with an edge structure to a real image. This paper deals with the stain slides of nuclei in tumor. We demonstrate several segmentation algorithms applied to the initial data set that contains real images and labels using synthetic augmentation in order to add their generalized images. We compute and report that a proposed synthetic augmentation procedure improves the accuracy indices.