A longitudinal study of the evolution of the central sulcus’ shape in preterm infants using manifold learning

Cortical folding in humans is different for every individual, and is associated with functional specificities. It forms mainly during the last trimester of pregnancy, hence its development lacks description, especially in a longitudinal way. To cope with this issue, this study focused on the evolution of the central sulcus’ variability of 71 preterm infants studied longitudinally with MRI at 30 and 40 weeks (w) postmenstrual age (PMA). Our aim was to investigate the main shape characteristics and whether they are encoded early on or appear closer to term birth. We captured shape dissimilarity between the sulci using a distance matrix after pairwise co-registration using an Iterative Closest Point algorithm. We applied non-linear dimensionality reduction to this matrix using the Isomap algorithm in order to capture the most discriminative shape features among the central sulci. We characterized the three most discriminative features over the group, and found that the sulci evolved consistently from a given feature at 30w PMA to the 40w PMA equivalent feature. We incidentally captured a feature that could coincide with the most discriminative adult feature, both visually and by its asymmetry in left and right sulcal distribution. These results captured the main shape features of the central sulcus in preterm infants and suggest that they are already encoded at 30w PMA.