Image Translation Based Nuclei Segmentation for Immunohistochemistry Images

AbstractNumerous deep learning based methods have been developed for nuclei segmentation H &E images and achieved close to human performance. However, direct application of such another modality images, as Immunohistochemistry (IHC) may not achieve satisfactory Thus, we a Generative Adversarial Network (GAN) approach translate an IHC image while preserving location morphology then apply pre-trained models the virtual image. We demonstrated that proposed work better than several baseline including state art Cellpose HoVer-Net, trained on generative method, DeepLIIF, using two public datasets.KeywordsGANH &EIHCNucleiSegmentation