Voxel importance 3D convolutional neural network using SHAP

Summary of what happened:

Hello everyone, I am trying to use shap library in python to do SHAP, a visualization technique that allows you to understand the behaviour of your model. Nevertheless, it is not prepared to 3D models so I can’t make it work.
Does anyone knows an implemented way to obtain SHAP values in 3D convolutional neural networks? I prefer python but I will take MATLAB or R etc…

Thank you!