Panoptic segmentation with highly imbalanced semantic labels
This work addresses the problem of accurately segmenting and counting colon nuclei in medical imaging, which is crucial for pathology and diagnosis, but it appears incremental as it builds upon existing architectures like Hovernet.
The authors tackled panoptic segmentation for colon nuclei identification by developing a weighted loss for imbalanced semantic labels and integrating it with a state-of-the-art instance segmentation model in a Hovernet-like architecture, achieving results for the CoNIC Challenge at ISBI 2022.
We describe here the panoptic segmentation method we devised for our participation in the CoNIC: Colon Nuclei Identification and Counting Challenge at ISBI 2022. Key features of our method are a weighted loss specifically engineered for semantic segmentation of highly imbalanced cell types, and a state-of-the art nuclei instance segmentation model, which we combine in a Hovernet-like architecture.