Towards Explainable Automated Neuroanatomy