Predicting the principal components of cortical morphological variables

Abstract

AO_SCPLOWBSTRACTC_SCPLOWThe generating mechanism for the gyrification of the mammalian cerebral cortex remains a central open question in neuroscience. Although many models have been proposed over the years, very few were able to provide empirically testable predictions. In this paper, we assume a model in which the cortex folds for all species of mammals according to a simple mechanism of effective free energy minimization of a growing self-avoiding surface subjected to inhomogeneous bulk stresses, to derive a new set of summary morphological variables that capture the most salient aspects of cortical shape and size. In terms of these new variables, we seek to understand the variance present in two morphometric datasets: a human MRI harmonized multi-site dataset comprised by 3324 healthy controls (CTL) from 4 to 96 years old and a collection of different mammalian cortices with morphological measurements extracted manually. This is done using a standard Principal Component Analysis (PCA) of the cortical morphometric space. We prove there is a remarkable coincidence (typically less than 8{whitebullet}) between the resulting principal components vectors in each datasets and the directions corresponding to the new variables. This shows that the new, theoretically-derived variables are a set of natural and independent morphometrics with which to express cortical shape and size.