Abstract
Shear Wave Elastography (SWE) imaging provides tumor stiffness information beyond conventional ultrasound (US) imaging for breast malignancy assessment. Unfortunately, the acquisition of SWE images is costly and labor-intensive, leading to severe data scarcity that hinders the training of discriminative models. Despite our best efforts, we were only able to collect around 550 SWE images. Fortunately, recent advances in diffusion-based generative models offer a promising solution to this limitation. However, synthesizing high-quality and high-fidelity SWE images from such limited data remains a challenge, as naively applying existing methods often leads to unsatisfactory results. To address this challenge, we propose US2SWEdiff, a ControlNet-based framework for generating realistic SWE images from corresponding US images. Using the generated synthetic datasets, we conduct comprehensive downstream experiments with evaluation on an external multi-center dataset. Results consistently demonstrate the superiority of US2SWEdiff over existing baselines, yielding a 10.220% improvement in AUROC and establishing a new state-of-the-art method for SWE image synthesis. Additionally, to facilitate future research, we release synthetic SWE image datasets generated from three public US datasets, BUSI, BUS-BRA, and Breast- Lesions-USG. Code is available at https://github.com/Jiaming21/US2SWEdiff.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/DGM4MICCAI_027.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/profile?id=~Jiaming_Huang1
BibTex
@InProceedings{HuaJia_Generation_MICCAISAT2026,
author = { Huang, Jiaming AND Qiu, Kunpeng AND Wang, Shuihua AND Liu, Feifei AND Yu, Kun-Hsing AND Bridge, Christopher P.},
title = { { Generation of Breast Tumors’ Shear Wave Elastography Images from Corresponding Ultrasound Images with US2SWEdiff } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17258},
month = {pending},
page = {pending}
}