Abstract
This challenge technical report addresses FU_Biometry, where one model localizes landmarks across prenatal, intrapartum, and cardiac ultrasound and is evaluated by both mean radial error (MRE) and biometric-parameter error. We present MABA, a measurement-aware biometry adaptation framework with a shared DINOv2 encoder, task-specific heatmap heads, continuous landmark decoding, measurement-aware objectives, gated pseudo labels, and manifest-level submission checks. On the refreshed 619-row visible-validation manifest, MABA reached an official CodaBench Average MRE of 24.2509 and task-level mean MAE of 30.90, compared with 32.2257 for the organizer baseline. A rejected FUGC-only probe scored 24.3607, showing that local transfer cues may not survive online validation. We therefore separate official scores, local proxy diagnostics, and single-student deployment claims.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/FoundUS_005.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=pEZs8QZ8xX
BibTex
@InProceedings{FanZha_MABA_MICCAISAT2026,
author = { Fan, Zhaorui AND Wang, Liwen AND Dong, Xingbo AND Jin, Zhe},
title = { { MABA: Measurement-Aware Biometry Adaptation for Landmark-Based Ultrasound } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17261},
month = {pending},
page = {pending}
}
