| Dexuan He | AI for Medicine |
AI for Medicine and Computational Pathology
I am Dexuan He, a postgraduate student at Shanghai Jiao Tong University working on AI for medicine, computational pathology, and spatial omics foundation models. My current research focuses on vision-language pathology models, whole-slide image analysis, rare cancer diagnosis, spatial omics, and clinically useful AI systems.
I am interested in building reliable AI systems that connect pathology images, biomedical knowledge, and clinical decision support. I also maintain public resources for computational pathology and release research artifacts when possible.
News
- Apr 2026: Our work on few-shot prompt tuning for rare cancer subtyping was published in Nature Communications.
- Apr 2026: Our knowledge-enhanced pathology vision-language foundation model was published in Cancer Cell.
- Apr 2026: We released KidRare, a whole-slide image dataset for rare pediatric pathology research.
Research Interests
- Computational pathology and whole-slide image analysis.
- Vision-language and multimodal foundation models for medicine.
- Spatial omics foundation models.
- Rare cancer diagnosis, subtype classification, and precision oncology.
- Medical knowledge integration and clinically meaningful AI evaluation.
- Robust and secure AI systems for biomedical signals.
Selected Open-Source Work
- PathPT: few-shot prompt tuning for rare cancer subtyping with pathology foundation models.
- KEEP: knowledge-enhanced pathology vision-language pretraining for cancer diagnosis.
- KidRare: a rare pediatric pathology WSI dataset released on Hugging Face.
- ManiBCI: code for Professor X, a robust backdoor attack study for EEG-based BCIs.
Collaboration
I am open to research collaborations on digital pathology, medical imaging, multimodal AI, rare cancer diagnosis, and AI for clinical trials. The best way to reach me is by email: firehdx233@sjtu.edu.cn.