Abstract

The neonatal brain undergoes rapid structural maturation during the perinatal period, yet accurate and anatomically interpretable assessment of brain maturity remains challenging. While deep learning–based brain age prediction models have shown strong predictive performance, most rely on post-hoc visualization techniques that provide limited region-level insight. In this study, we propose an interpretable neonatal brain age prediction framework guided by a pretrained model for neonatal brain segmentation. The framework extracts region-wise representations from a pretrained segmentation model and integrates them through a multi-region cross-attention module, enabling quantitative analysis of anatomical contributions to age prediction. Experiments on the Developing Human Connectome Project dataset and an independent institutional dataset demonstrate that the proposed model outperforms direct structural image-based brain age prediction baselines. Region-level analysis reveals age-dependent shifts in anatomical contributions, characterized by increasing gray matter importance and decreasing basal ganglia and ventricular contributions with advancing age. These findings suggest that segmentation-guided region-wise attention may improve interpretability in neonatal brain age prediction.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/PIPPI_020.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to Open Review

Open Review Page: https://openreview.net/forum?id=WB96IsKIwt&referrer=%5BProgram%20Chair%20Console%5D(%2Fgroup%3Fid%3DMICCAI.org%2F2026%2FWorkshop%2FPIPPI%2FProgram_Chairs%23submission-status)

BibTex

@InProceedings{BakDay_RegionWise_MICCAISAT2026,
        author = { Bak, Dayeon AND Kang, Junghwa AND Kim, Hyun Gi AND Nam, Yoonho},
        title = { { Region-Wise Interpretable Neonatal Brain Age Prediction via Pretrained Segmentation-Guided Attention } },
        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}
}


back to top