Clarification on Feature Extractor Used for MIL Baselines in Table 1

[thomascong121]

Thank you for the great work and for sharing the code.

I was wondering if you could clarify what feature extractor was used for patch-level feature extraction in the MIL baselines reported in Table 1? Specifically, was a standard backbone (e.g., ResNet50 pretrained on ImageNet) used, or did you employ a domain-specific or self-supervised model (like CONCH, the same one used for Madeleine)?