Adversarial Nuclei Segmentation on H&E Stained Histopathology Images

Computer aided methods in pathology are advancing rapidly. Problems like segmentation, classification and detection of pathology images are solved with machine learning and image processing techniques. State-of-the-art methods in nuclei segmentation problem include supervised deep learning techniques. However, labeling process of pathology images is an expensive and time consuming process. In this work, nuclei segmentation problem is formulated as image-to-image translation problem and using Cycle-Consistent Generative Adversarial Networks, an unsupervised segmentation scheme is proposed for hematoxylin & eosin stained histopathology data.