About Pathology Feature extraction shape

Cassie07 / PathOmics

[MICCAI 2023 Oral] The official code of "Pathology-and-genomics Multimodal Transformer for Survival Outcome Prediction" (top 9%)

72 stars 10 forks source link

About Pathology Feature extraction shape #10

Closed zhengbiqing closed 1 month ago

zhengbiqing commented 4 months ago

Thank for your great work! I have a question and please give me a help.

Different WSI has different size and will be cut in different amount small images (for example 256256). Each small image correspond to a feature (for example, as a tensor of 11024). So different WSI has different shape feature (for example, N1024, M1024). How does these different shape features as input to model and train the model?

Looking forward to your answer. Thanks very much!

Cassie07 commented 3 months ago

Hi @zhengbiqing, Thanks for your interest in our study! We set the batch size to 1. So when you load the data by data loader, you do not have to force the number of patches to be the same for each patient. Please feel free to let me know if you have any other questions.