Abstract
Fetal movement is an important indicator of fetal well-being and underpins routine obstetric ultrasound (US) tasks such as biometric assessment and standard-plane acquisition. Accurate motion estimation is therefore essential for both clinical analysis and autonomous US systems. However, anatomical changes observed in freehand US image sequences arise from the combined effects of fetal and probe motion, creating a motion ambiguity that makes fetal-motion estimation inherently ill-posed. To address this challenge, we introduce FetalSense, a framework for disentangling fetal and probe motion from US image pairs. FetalSense is trained using simulation-derived SE(3) pose supervision for both the fetus and probe, while requiring only US images and tracked probe motion at inference. We further introduce Anatomical Landmark Displacements (ALDs), an image-derived signal of fetal motion based on temporal changes in fetal anatomy, thereby providing supervision without requiring fetal-pose measurements at deployment. FetalSense combines a probe-pose-conditioned image encoder with an ALD-guided landmark stream to enable explicit motion decomposition. We evaluate the method on a large-scale fetal US motion dataset generated using our simulation framework, which will be released publicly upon acceptance. FetalSense achieves a mean fetal displacement error of 1.56 mm and a decomposition closure error of 0.69 mm. The learned fetal-motion representation further supports motion-state classification, achieving 94.1% accuracy and 96.7% F1-score for significant fetal-movement detection. By jointly introducing fetal–probe motion disentanglement and ALDs as an image-derived motion cue, FetalSense establishes a foundation for motion-aware fetal US analysis and autonomous US navigation.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/RIME_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=Yxe9zOp1Hc
BibTex
@InProceedings{BalSow_FetalSense_MICCAISAT2026,
author = { Balaji, Sowjanya AND A., Anusha AND Ram, Keerthi AND Ayyasamy, Shyam AND Seshadri, Suresh AND Lakshmanan, Manojkumar AND Sivaprakasam, Mohanasankar},
title = { { FetalSense: Anatomy-Guided Fetal–Probe Motion Disentanglement in Freehand Ultrasound } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17277},
month = {pending},
page = {pending}
}