Distill PathologyDistill Pathology
How should a pathology distillation study test efficiency, external validity, and calibration before making deployment claims?病理知识蒸馏研究应如何检验效率、外部有效性与校准，才能避免过早作出部署论断？
Why this project exists为什么开展这个项目
A proposed study of knowledge distillation from a pathology foundation model to a lightweight classifier.一项把病理基础模型知识蒸馏到轻量分类器的拟议方法研究。
The falsifiable question可证伪的问题
How should a pathology distillation study test efficiency, external validity, and calibration before making deployment claims?病理知识蒸馏研究应如何检验效率、外部有效性与校准，才能避免过早作出部署论断？
How the question is tested如何检验这个问题
- Proposed DINOv2 teacher → lightweight CNN student distillation拟议的 DINOv2 教师模型 → 轻量 CNN 学生模型蒸馏
- Planned cohort-separated external validation计划按队列分离进行外部验证
- Planned calibration and uncertainty analysis计划开展校准与不确定性分析
What the current evidence supports现有证据支持什么
No quantitative result is displayed here until a verified result and its evidence path are ready for public release. The project description therefore records scope and method, not a performance claim.经核实的结果和它的证据路径可以公开之后，本页才会展示定量结果。所以这里记录的是范围和方法，不是性能论断。
Where the claim stops论断止于何处
- No dataset has been selected and no model has been trained or evaluated.目前尚未选择数据集，也尚未训练或评估模型。
- The page describes a study design and makes no performance or clinical-use claim.本页只描述研究设计，不作性能或临床使用论断。
What comes next接下来做什么
- Select a licensed dataset and preregister the evaluation protocol before implementation.选择许可清晰的数据集，并在实施前预先确定评估方案。