Abstract
In this preliminary method paper, we describe our proposed approach for the MICCAI 2026 Foundation Model for Ultrasound Biometry challenge. This challenge demands a unified, generalizable framework to detect anatomical landmarks and replicate clinical biometric measurements across ten tasks covering prenatal, intrapartum, pediatric, and adult cardiac ultrasound data collected from multiple centers and diverse devices. Each task includes a small annotated subset and a larger pool of unannotated data, partially sourced from videos. Our method integrates three complementary strategies. First, we explore whether pretraining the backbone via LeJEPA on the unlabeled frame corpus produces an encoder more adapted to ultrasound imaging than a DINO-based backbone. Second, we leverage unlabeled data with pseudo-labeling by combining optical-flow propagation and curriculum pseudo-labeling: we expand the training dataset starting from sparse ground truth, propagating labels first along the temporal dimension of source videos and then across the whole dataset based on embedding similarity. Third, we adopt an anatomy-aware post-processing module to regularize raw predictions with task-specific training statistics and geometric priors tailored to each measurement, including measurement orientation, landmark laterality, and the anticipated intensity profile of target anatomical structures.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/FoundUS_122.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: Not Available
BibTex
@InProceedings{RuiJon_PULSE_MICCAISAT2026,
author = { Ruiz-Molsgaard, Jonatan AND Ibragimov, Bulat},
title = { { PULSE: Pseudo-labelling of Ultrasound by Local-to-Semantic Expansion for Generalisable Biometry } },
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}
}