Unsupervised Nuclei Segmentation Using Spatial Organization Priors

In digital pathology, various biomarkers (e.g., KI67, HER2, CD3/CD8) are routinely analyzed by pathologists through immuno-histo-chemistry-stained slides. Identifying these on patient biopsies allows for a more informed design of their treatment regimen. The diversity and specificity types images make the availability annotated databases sparse. Consequently, robust efficient learning-based diagnostic systems difficult to develop apply in clinical setting. Our study builds observation that overall organization structure observed tissues similar across different staining protocols. this paper, we propose leverage both wide haematoxylin-eosin stained invariance tissue order perform unsupervised nuclei segmentation immunohistochemistry images. We implement evaluate generative adversarial method relies high-level distribution priors comparison with largely available cell masks. approach shows promising results compared classic supervised methods, as quantitatively demonstrate two publicly datasets. code is encourage further contributions ( https://github.com/loic-lb/Unsupervised-Nuclei-Segmentation-using-Spatial-Organization-Priors ).