This model is an updated version for neonatal T1w images, and is trained based on HippUnfold SynthSeg v0.1 segmentations, which were manually QC's and corrected by the Duerden Lab. The training workflow is in https://github.com/khanlab/hippunfold-train-finetune/tree/downgrade-v1. Note that the nnUNet training used nnUNetTrainer instead of nnUNetTrainerV2 (inadvertently), so a minor fix was needed in hippunfold to parse and pass along the trainer from the tar file. (not re-running since it was 5*3 GPU-days to compute..