Human connectome: Hierarchical clustering

Building a structural connectome involves (1) subdivision of the brain’s grey matter into functio-anatomically relevant areas and (2) estimation of the connectivity between them. With respect to (1), there is ample evidence that there are distinct brain areas with respect to certain structural criteria, e.g., cytoarchitecture. However, any particular parcellation provides an incomplete view on the brain’s functional-anatomical organization. First, the parcellation depends on the specific criteria used to define similarity. It makes much sense to use connectivity as parcellation criterion, as brain function is supported mainly by networks on different scale levels – from the local circuits within a microcolumn to brain-wide networks mediated by white matter fibers. It has been demonstrated that cortex parcellations based on long-range connectivity is meaningful and reproducible. Second, the parcellation depends on the desired granularity to describe the similarity structure of the brain. Here, we will present connectivity based hierarchical parcellation techniques that yield representations of entire families of parcellations. Concerning (2), there are various techniques to establish anatomical connectivity (e.g., tracing or dissection), but in living humans, only diffusion MRI based techniques are applicable. Here we will use advanced techniques for diffusion tractography based on high angular resolution diffusion imaging (HARDI) data