#### Weakly-Supervised semantic segmentation for histopathology images based on dataset synthesis and feature consistency constraints
---
- homogenity of medical images
---
### WSSS for medical images
- multiple instance learning (MIL)
- CAM-based methods
- HistoSegNet (GradCAM)
- WSSS-tissue
- C-CAM (MRI)
- OEEM
---
- **Problems:** though the above methods can alleviate fuzzy boundary problem to some extent, the gap between classification and segmentation still exists
---
- PistoSeg framework
---
- pseudo-mask refining module
---
- Loss design
---
- performance on WSSS4LUAD / BCSS-WSSS
---
- ablations on losses
---
- ablations on refining module on different CAMs
---
- ablations on precise segmentation inputs
---
---
{"title":"Weakly-Supervised semantic segmentation for histopathology images based on dataset synthesis and feature consistency","description":"type: slide","contributors":"[{\"id\":\"4724c2c9-74e7-418c-87c3-79a619b97461\",\"add\":1477,\"del\":95}]"}