Do We Need a Human post mortem Whole-Brain Anatomical Ground Truth in in vivo Magnetic Resonance Imaging?

Non-invasive in vivo neuroimaging techniques provide a wide array of possibilities to study human brain function. A number of approaches are available that improve our understanding of the anatomical location of brain activation patterns, including the development of probabilistic conversion tools to register individual in vivo data to population based neuroanatomical templates. Two elegant examples were published by Horn et al. (2017) in which a method was described to warp DBS electrode coordinates, and histological data to MNI-space (Ewert et al., 2017). The conversion of individual brain scans to a standard space is done assuming that individual anatomical scans provide a reliable image of the underlying neuroanatomy. It is unclear to what extent spatial distortions related to tissue properties, or MRI artifacts exist in these scans. Therefore, the question rises whether the anatomical information from the individual scans can be considered a real ground truth. To accommodate the knowledge-gap as a result of limited anatomical information, generative brain models have been developed circumventing these challenges through the application of assumption sets without recourse to any ground truth. We would like to argue that, although these efforts are valuable, the definition of an anatomical ground truth is preferred. Its definition requires a system in which non-invasive approaches can be validated using invasive methods of investigation. We argue that the application of post mortem MRI studies in combination with microscopy analyses brings an anatomical ground truth for the human brain within reach, which is of importance for all research within the human in vivo neuroimaging field.