fMRI RSA: how to treat different sessions of the same subject

Hi experts,

I am running an RSA analysis and I have a question. After the GLM in SPM, I have one beta image for each regressor for each run.

My current approach was to calculate a RDM for each run and then doing an average of these matrices to get one matrix per subject. Is this reasonable or there is a better way of dealing with this?

Another related question: before computing the RDM, should I normalise the beta images within or across runs?