About how the .pt file after IG-SG is made interpretable is mapped back to the standard brain and thus visualised

[katouMegumiH]

Here's the thing, the final output feature matrix in the model dataset is not the original standard brain dimensions due to the change in dimensions, so I'm asking how the interpretable plots in your paper were drawn? How do you map the IG-SG output .pt file back to the standard brain? What tool was used for the final drawing? The highlighted areas in the figure are very nice, when I visualised it, I first did the inverse of the feature matrix output from IG-SG as in dataset with the feature dimension transformation, and then visualised it with nilearn, but the visualised figure is not a bunch of dots instead of the highlighted areas.

[Padraig20]

Hi @katouMegumiH,

thanks for your question! You should check out this tutorial on SwiFT. You will find a section called Predict sex with pre-trained models and interpret with Integrated Gradient, where you will see a direct example on how you can interpret your results after training!

Hope this helps, let me know :)