Gyral Growth Patterns of Macaque Brains Revealed by Scattered Orthogonal Nonnegative Matrix Factorization

Abstract

Cerebral cortex development undergoes a variety of alternate processes, providing valuable information to study the developmental mechanism of the cortical folding and the structural and functional architectures. Many longitudinal studies are performed on the development of sulci using features like sulcal depth, but the gyral system is less studied. To fill the gap, we propose a novel feature, termed gyral height, to quantify the longitudinal developmental patterns of gyri. Another practical problem is the difficulty of obtaining data for all time points for all subjects, even in animal datasets, such as the macaque neurodevelopment dataset in this work. Therefore, we develop a novel method by introducing a scattered factor to the orthogonal nonnegative matrix factorization to align data both longitudinally and cross-sectionally. By this method, the gyral height feature maps are decomposed into orthogonal cortical clusters which encode spatiotemporal patterns. Close relations are found between these clusters and anatomical, structural connective and functional metrics, suggesting the potential of the novel cortical feature and the method in investigating the brain development.