Classification options

[jeroenmollink]

Hi Richard,

Thanks for sharing this code, it looks very interesting. Is possible to train PatchGCN_Surv in a classification manner? For example, can we use the hazards variable to model the risk probabilities of a WSI and use these as input for cross entropy loss to train the model?

Thanks,
Jeroen

[Richarizardd]

Hi @jeroenmollink - Thank you for taking interest in our code! It is possible to train Patch-GCN in a classification manner. You would simply need to modify the output dimension of the last layer to be the # of classes (See the following code snippet for CLAM, which does classification.)

Performance of Patch-GCN may vary on other tasks. As described in the main paper, the intuition behind Patch-GCN was largely motivated by the next of context-aware features in survival (beyond simple instance-level features). For simple needle-in-a-haystack tasks, Attention MIL may still be sufficient in a lot of scenarios.