TRIAGE-MIL
Multi-axis instance selection and semantic hypergraph modeling for WSI survival prediction
TRIAGE-MIL is a whole-slide image survival prediction project that combines multi-axis instance selection and semantic hypergraph modeling for patient-level risk prediction from histopathology images.
The project aims to improve survival prediction by selecting informative WSI instances and modeling higher-order semantic relationships among tissue regions.
Role: Lead researcher
Key contributions:
- Developed a survival prediction framework for whole-slide images.
- Designed multi-axis instance selection for identifying informative image regions.
- Integrated semantic hypergraph modeling to capture relationships among selected WSI instances.
- Evaluated the method for patient-level survival prediction from histopathology data.
- Prepared the manuscript for MICCAI 2026 submission.
Research focus:
- WSI survival prediction
- Multiple instance learning
- Hypergraph modeling
- Patient-level risk prediction
- Computational pathology