Status: Currently our DCM seg writer skips empty slices (thats the default of dcmqi). This results in the fact, that segs stored and loaded into MITK cannot be directly used for e.g. statistics or modifying the segmentation, because we assume/test in most code parts that mask and image have the same geometry, which will mostly not be the case (or in other words MITK currently does not support yet segmentations that are only a cropped grid of the image, so smaller extend but same geometry).
I think we should at least change to not skip the empty slices on save for now. To produce segs that MITK fully supports currently. If we support cropped segmentations somewhen in the future, we could switch it on again.
This seems also to mitigate the problem noted in T26949.