Weakly Supervised Cell Instance Segmentation for Multi-Modality Microscopy

Instance segmentation of multi-modality high-resolution microscopy images is an important task in computational pathology. We extended HoVer-Net[1], originally developed for segmentation and classification of nuclei in multi-Tissue histology images, to apply it under weakly supervised situation. According to the final tests, this modification also works for multi-modality microscopy.