I am aiming for a longitudinal volumetric analysis of a clinical paediatric dataset (mostly around 5 timepoints with a 3D T1w) based on freesurfer. Unfortunately, during the acquisition period a change of the acquisition protocol occurred with changes in terms of TE/TR, voxel and dimension size. This resulted in nearly all subjects having data of both acquisition protocols. It’s an interventional study, hence we are interested in the baseline scans pre-intervention (mostly protocol 1) and how intervention alters the trajectories of brain volume changes. Now, a substantial share of the post-intervention scans were acquired with protocol 2. We appreciate the problem of the systematic difference between scans flawing potential results. Due to a quite small sample size and we want to avoid discarding any data.
I know of efforts like Combat to address harmonization of multi-site data. However, as this is a quite special case of within subject protocol heterogeneity I am not sure whether I can apply it in our analysis? Are there any other tools or strategies you recommend?
Grateful for any input and guidance.