Abstract
Breast ultrasound (BUS) videos provide essential temporal cues but suffer from dense annotation scarcity, whereas static images offer stable morphology without motion context. We propose ProME-Net, an asymmetric heterogeneous fusion framework for BUS video detection. Specifically, an Anchor Frame aggregates temporal context to mitigate probe-induced artifacts, while a distinct Prototype Frame extracts reusable structural priors learned from a mixed static-video stream. Driven by clinical screening demands, we prioritize high-recall precision and false positives per image (FP) in our evaluation. On a mixed-dimension benchmark under a dual-task setting, ProME-Net achieves 62.6% $AP_{50}$, 49.0% $Pr_{90}$, 0.685 $FP_{90}$, and 58.72 FPS, demonstrating superior false-positive control without compromising overall accuracy.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/ASMUS_018.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=ayza8oJOzt
BibTex
@InProceedings{WuZhe_ProMENet_MICCAISAT2026,
author = { Wu, Zhengzheng AND Tao, Xing AND Zhang, Yuhan AND Xu, Jinning AND Si, Su AND Ni, Dong},
title = { { ProME-Net: Prototype-Guided Artifact-Robust Breast Ultrasound Video Detection } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17276},
month = {pending},
page = {pending}
}