CTranspath encode patch size

[HongLiuuuuu]

Thank you for your excellent work! I noticed that you use CTranspath to encode raw image patches into feature vectors. The patches are extracted at 20x magnification with dimensions of 256 × 256. However, CTranspath requires patches of 224 × 224. I’m curious to know how you addressed this?

[ajv012]

Thanks for your question and interest in our work! We resized the 256x256 patches to 224x224 before feeding them to CTransPath.

When this work was done, strong patch encoder foundation models were not widely available. We invite you to try out SurvPath with newer models like UNI, GigaPath, and Virchow.

Please let us know if you have any additional questions.

[HongLiuuuuu]

Thank you for your quick response! I was wondering, if we directly resize 256 × 256 patches to 224 × 224, does the magnification remain at 20x?