Abstract
Magnetic Resonance Imaging (MRI) provides a non-invasive means to examine pediatric brain anatomy. However, in low-resource and point-of-care settings, ultra-low-field scanners are increasingly used because of their affordability and portability, while their low signal-to-noise ratio and weak tissue contrast make accurate segmentation of small sub-cortical and limbic structures challenging. In this work, we present our LISA2026 Task2 submission for eleven-structure segmentation from 0.064T pediatric T2-weighted MRI. Built on a residual-encoder (Res-Enc) nnU-Net backbone, the system combines low-field intensity representations, anatomical structure priors, global/local feature fusion and intensity-conditioned decoder modulation. To improve robustness under limited labelled data, we combine four configured members that target complementary failure modes related to image context, anatomical localisation and shape-aware optimisation. On the official LISA 2026 validation set, the submitted ensemble achieves a mean Dice similarity coefficient (DSC) of 0.82, mean 95th-percentile Hausdorff distance (HD95) of 1.71 and mean average symmetric surface distance (ASSD) of 0.47. The code is publicly available at: https://github.com/benzxcvasdf/
LISA2026_3590290.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/LISA_010.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=cTGsOWtSNl
BibTex
@InProceedings{LamKa_Generalized_MICCAISAT2026,
author = { Lam, Ka-Ho AND Lin, Pei-Jung},
title = { { Generalized Prior-Conditioned MRI Segmentation for Ultra-Low-Field Pediatric Brain Structure Segmentation } },
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}
}