Hi Paul,
Thanks for this. We can see where the error is arising - for at least one voxel the number of neighbouring voxels available for consideration in the spatial VB step is zero even though that voxel should be considered as one of its own neighbours.
We're really not sure what about your data could cause this. Would you be willing to share one of the datasets with us so we can have a look at it?
All the best,
Flora