Hi,
I have been working with a set of images our research group gathered. We tried doing an SPM analysis, but didn’t get any significant pattern. After reviewing the studies, I noticed that we have 10 different acquisitions:
We tried to work around how to harmonize this, since the images are notably different once normalized to MNI space, as it is possible to see that the smoothing factor differs. Additionally, since we only had CT and PET images for the majority of the studies, I compared different normalization methods applied to PET images (unified normalization by SPM on the ones with MRI, Unified spatial normalization method of brain PET images using adaptive probabilistic brain atlas, and synthmorph), and the one that got the best mutual information score with Della Rosa template was Synthmorph.
We now try to harmonize the data. According to this papper , in order to harmonize the studies due to the acquisition parameters, the phantom calibration images are mostly needed. It was not possible to gather this data, and I noticed that the ones I could gather were not scanned by any of the previous protocols. We have proposed to work on a data-driven approach to harmonize these studies, but couldn’t advance much more.
If someone has dealt with this situation before and has an idea, I would like to discuss.
Thank you.