Surface-based subfield alignment pipeline. A) Each 3D dataset manually parcellated into subfields. B) Subfields mapped to a common unfolded space using HippUnfold. C) Morphometric features mapped to unfolded space (from left to right: gyrification, thickness, curvature). D) Multimodal features iteratively aligned in unfolded space using 2D registration. E) Subfield labels propagated through unfolded 2D registrations to sample1. Closeups of the segmentations for the individual datasets can be found in Supp. Figure 2.

Control condition using volumetric registration. A) Subfield segmentation as in Figure 1A. B) Binarized and rigidly aligned hippocampal gray matter masks (left hemispheres flipped). C) Iterative alignment using ANTs template building in 3D. D) Each sample’s subfields propagated to template space. E) Each sample’s subfields propagated to sample1 unfolded space.

Evaluation of aligned subfield definitions. A) Qualitative example of subfields from the third sample projected to the first sample’s native “folded” space using conventional 3D volumetric alignment, unfolding to account for inter-individual differences in folding shape, and unfolding followed by registration in unfolded space. B) Dice overlap achieved. Each measure was calculated in unfolded space (left) and again in the first sample’s (BigBrain left hemisphere) native folded space (right). Black lines indicate the mean across all subfields. C) Distances between all aligned subfield borders using the three methods described above. Dashed lines indicate the median distance.

Contribution of each morphological feature to unfolded registration performance. A) Unfolded space registration was repeated for all combinations of unfolded morphological features and evaluated by Dice overlap in native space. Combinations are ordered by their Dice scores averaged across the 5 subfields. p-values are relative to no unfolded registration, using one-tailed paired-samples t-tests as above. B,C) Evaluation of which hippocampal vertices (B) and subfields (C) were most deformed in unfolded registration.