Abstract
Accurate choroid plexus (CP) segmentation in neonatal MRI is important for assessing CP volume and morphology. Reliable CP measures may provide insight into cerebrospinal fluid dynamics, glymphatic function, and early neurodevelopment. However, CP segmentation in neonates remains challenging due to rapid age-related changes in morphology and tissue contrast. In this work, we propose an automated age-conditioned framework for CP segmentation in neonatal T1-weighted MRI from the developing human connectome project dataset (train n=69; tuning n=12; validation n=19; test n=40). The method uses a 3D U-Net with conditional instance normalization, where postmenstrual age (PMA) at scan is encoded and used to modulate feature distributions. To reduce reliance on manual layer selection, we introduce a rule-based controller that monitors layer-wise contribution during training and dynamically identifies age-sensitive layers. The method was evaluated on neonatal MRI and compared with U-Net, sub-age-restricted U-Net, age-concatenation U-Net, FiLM U-Net, CIN U-Net, and nnU-Net. The proposed method achieved a Dice similarity coefficient of 0.85 ± 0.08, compared with 0.73–0.81 for the comparison methods. Boundary metrics were also improved, with a 95th-percentile Hausdorff distance of 2.80 mm and an average symmetric surface distance of 0.59 mm (paired Wilcoxon, p<0.05). Age-stratified analysis showed improved performance across most PMA groups examined, including the youngest infants, and layer-wise analysis indicated larger negative ∆Dice values in decoder stages. These results suggest that self-tuning age conditioning may support neonatal CP segmentation and developmental studies requiring reliable CP quantification across early brain maturation.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/PIPPI_017.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
BibTex
@InProceedings{KanJun_SelfTuning_MICCAISAT2026,
author = { Kang, Junghwa AND Bak, Dayeon AND Kim, Hyun Gi AND Shin, Na-Young AND Nam, Yoonho},
title = { { Self-Tuning Age-Conditioned Choroid Plexus Segmentation in Neonatal MRI } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17257},
month = {pending},
page = {pending}
}