Hi Mike, hi Marcelo,
According to my test, here the error “Inputs do not occupy the same physical space!” was induced due to insufficient memory (to me, this is really weird). The last two steps of the (ANTs Syn – subcortical refine) normalization require significant amount of memory. I also saw this error when I tested on my pc with 16GB memory. But there’s no problem when I tested it on our server with very large memory.
Actually, the registration is already done when the error happens. But the memory is not released somehow. ANTs will throw out the error when further generating the composite transforms.
This error doesn’t happen when using ANTs Syn normalization without subcortical refine.
In short, could you maybe test it on a machine/cluster with above 24GB (or better 32GB) memory to see if this error still exists? I will also check the parameter setting and ANTs code to see if there’s any clue on this.