ROI in the data is too large for SVoRT about nesvor

[daviddmc]

Hi @m-dannecker,

Thanks for being interested in our work. Since the SVoRT model is trained on fetal brain data, it will throw a warning if the input ROI is too large. However, it should still perform the registration, and you may ignore the warning if you want to use SVoRT. Also, for brain data that is surrounded by air, e.g., adult brain and neonatal brain, it is recommended to remove the air background first, for example, using --otsu-thresholding.

[m-dannecker]

Thanks for the quick reply!
Good to know. I guess the stack transformation provides a better metric then, as it always proceeds with stack transformation. Regarding air around the brain: the HCP brains come pre-masked, so that should not be a problem, right?

[daviddmc]

By default, the command runs both SVoRT and stack-to-stack registration and chooses the method with higher NCC consistency.
If the interslice misalignment is relatively small, it would prefer stack registration.

[m-dannecker]

Alright, thanks for the clarification!