We rely on preprocessing to ensure a minimum image size. If collaborators failed to run the preprocessor before executing the model code, they get a RuntimeError from torchio.
We could catch this error, and if it is the patch-size issue, instruct them to run pre-processing.