What would you like to see added in this software?

For PET, dynamic frames in the beginning of a sequence have very little SNR, and so this is a very difficult registration problem. The same goes potentially for frames at the end of a sequence, where low SNR frames also might be difficult to register. To accommodate this, it would be ideal if we could choose a given timepoint for the modeling/registration to start, and then the transforms/motion estimated for this first frame, will be subsequently be applied to the remaining frames prior to this timepoint. This will lead to more stable outputs.

Do you have any interest in helping implement the feature?

Yes

Additional information / screenshots

This will avoid these high-peak incorrect motion estimations and subsequent registrations (frame 10 is at 2 minutes), and in previous PET motion correction workflows we have used the 2 minute mark as the first frame to register/estimate motion, and then assuming that the participant is lying still in the first 2 minutes of a scan.