Developing Human Connectome Project spatio-temporal surface atlas of the fetal brain

This repository contains spatio-temporal surface atlas, spanning 21-36 weeks of gestation, generated by adapting the procedure developed for the neonatal surface atlas in Bozek et al., 2018, which iteratively refines templates through a repeated alignment of individuals to a common space using MSM algorithm to the fetal data. At each iteration, the age-specific templates are obtained through weighted averaging of co-registered surfaces, whereby weights are defined by the Gaussians centred on the gestational weeks for which templates are calculated, and are subsequently used as a target space for the following iteration. Adaptive kernel regression, compensating for a difference in the number of scans available for different ages, was used to parameterise the width of the Gaussians. The atlas was generated in three stages: first, a common reference space was initialised via affine sulcal-depth-based registration to the dHCP neonatal GW36 template. At the next iteration, the template was refined using sulcal-depth-based nonlinear alignment, followed by 4 iterations of curvature-based alignment (a more fine-grained feature than the sulcal depth).