Distinct structural patterns of the human brain: A caveat for registration

Frithjof Kruggel

Keywords: structural patterns, connectivity, human cortex

Paper Status: original work, not submitted yet

TL;DR: We demonstrate here that structural variants of brain regions with distinctive properties exist in a population.

Abstract: Cortical structures of the human brain show a puzzling complexity and inter-individual variability. Numerous analytic approaches implicitly assume that structural properties of brains, represented in any high-dimensional space, form a single cluster and use nonlinear registration to reduce the inter-individual variability. We challenge this assumption. Depending on the features and similarity criteria involved in the registration process, the total variance is reduced by only 20-40%. Consider a simplifying analogy: Suppose we want to study structural properties of cars. We hardly doubt that a registration procedure can be designed that successfully matches gross car parts (e.g., the passenger and engine compartment, the trunk and wheels). However, when zooming into details, objects under study become distinct (e.g. a trunk of a truck vs. a sports car, a combustion engine vs. an electric motor). Here, we demonstrate here that structural variants of brain regions with distinctive properties exist in a population. Avoiding an arguable registration and embracing the actual variability leads to analytic procedures that actually explain sources of variability at a considerably larger proportion.