How to use Heatmap for My model?

[Tengxf]

Hello, thank you very much for sharing your code. I’ve encountered an issue while trying to visualize the attention scores from my model.

The attention scores are stored in a variable attn, which is a np.ndarray with a shape of (4643,). The values range from [0.019, 1.0], with an average of around 0.192. I want to visualize them on the lowest-level WSI. My data consists of two WSIs—one at 20X magnification and the other at 40X. After averaging the attention scores from my model for these two magnifications, I use them as the scores for the heatmap, while the coords I pass are the patch coordinates from the 20X WSI. However, after running the code, the visualization doesn’t show the regions of attention—instead, I only see either the original image background or a completely blank image.

Could you please advise on what might be causing this issue? I’ve been debugging for a while but haven’t been able to successfully visualize the attention. I’d really appreciate your help. Looking forward to your reply—thank you very much!

[guillaumejaume]

Consider checking our new sister repo Trident, in particular this tutorial: https://github.com/mahmoodlab/TRIDENT/blob/main/tutorials/3-Training-a-WSI-Classification-Model-with-ABMIL-and-Heatmaps.ipynb.

If you already have a pretrained MIL model with a common patch encoder, it should be pretty straightforward.

[Tengxf]

请考虑检查我们的新姊妹存储库 Trident ，特别是本教程： https://github.com/mahmoodlab/TRIDENT/blob/main/tutorials/3-Training-a-WSI-Classification-Model-with-ABMIL-and-Heatmaps.ipynb。

如果您已经有一个带有通用贴片编码器的预训练 MIL 模型，那么它应该非常简单。

Thank you. I have already solved this problem now. Thanks for your reply.