A Semi-Supervised Framework for Breast Ultrasound Segmentation with Training-Free Pseudo-Label Generation and Label Refinement

Ruili Li; Jiayi Ding; Ruiyu Li; Yilun Jin; Shiwen Ge; Yuwen Zeng; Xiaoyong Zhang; Eichi Takaya

cs.CVMar 6, 2026

A Semi-Supervised Framework for Breast Ultrasound Segmentation with Training-Free Pseudo-Label Generation and Label Refinement

Authors: Ruili Li, Jiayi Ding, Ruiyu Li, Yilun Jin, Shiwen Ge, Yuwen Zeng, Xiaoyong Zhang, Eichi Takaya,

Organizations: Tohoku University Graduate School of Medicine, Sendai 980-8575, Japan · State Key Laboratory of Oncology in South China, Sun Yat-sen University Cancer Center, Guangzhou 510060, China · Southeast University, School of Cyber Science and Engineering Nanjing, Jiangsu, China · School of Software Technology, Zhejiang University, Ningbo 315100, China ·

Abstract

Semi-supervised learning (SSL) has emerged as a promising paradigm for breast ultrasound (BUS) image segmentation, but it often suffers from unstable pseudo labels under extremely limited annotations, leading to inaccurate supervision and degraded performance. Recent vision-language models (VLMs) provide a new opportunity for pseudo-label generation, yet their effectiveness on BUS images remains limited because domain-specific prompts are difficult to transfer. To address this issue, we propose a semi-supervised framework with training-free pseudo-label generation and label refinement. By leveraging simple appearance-based descriptions (e.g., dark oval), our method enables cross-domain structural transfer between natural and medical images, allowing VLMs to generate structurally consistent pseudo labels. These pseudo labels are used to warm up a static teacher that captures global structural priors of breast lesions. Combined with an exponential moving average teacher, we further introduce uncertainty entropy weighted fusion and adaptive uncertainty-guided reverse contrastive learning to improve boundary discrimination. Experiments on four BUS datasets demonstrate that our method achieves performance comparable to fully supervised models even with only 2.5% labeled data, significantly outperforming existing SSL approaches. Moreover, the proposed paradigm is readily extensible: for other imaging modalities or diseases, only a global appearance description is required to obtain reliable pseudo supervision, enabling scalable semi-supervised medical image segmentation under limited annotations.

Figures & tables

Fig. 1: To illustrate how different textual prompts affect zero-shot knowledge transfer, we visualize bounding box generated from (c) medical terms (“tumor”), (d) radiological attributes (“high density”), and (e) appearance-based descriptions(“dark oval.dark round.dark lobulated”).

Fig. 2: Overview of the proposed semi-supervised BUS segmentation framework. The pipeline consists of two stages: (1) Appearance-Prompted Pseudo-Label Generation (APPG), where appearance-prompted vision–language models (VLMs) produce initial pseudo labels in a training-free manner; (2) Pseudo-Label Refinement, which includes two steps: static-teacher warm-up training to capture coarse structural priors of breast lesions, and uncertainty-based semi-supervised learning using a dual-teacher framework with Uncertainty–Entropy Weighted Fusion (UEWF) and Adaptive Uncertainty-Guided Reverse Contrastive Learning (AURCL).

Fig. 3: Structure of the proposed AURCL module. Entropy maps identify top- $k$ high-uncertainty pixels, whose predictions are reversed and aggregated into patch-level features. Contrastive learning then pulls the original and reversed high-entropy features from these high-entropy patches closer to each other, while pushing them apart from stable low-entropy patches.

Fig. 4: Visual comparison of different state-of-the-art methods on the BUSI and UBB datasets. All models are trained in 2.5% labeled data. Red, green and yellow regions represent ground truth, prediction and overlapping regions, respectively.

Method	Venue	BUSI					UBB (UDIAT + BREASTUSG + BUSUCLM)
Method	Venue	Labeled	Unlabeled	Dice(%) $\uparrow$	IoU(%) $\uparrow$	Acc(%) $\uparrow$	Labeled	Unlabeled	Dice(%) $\uparrow$	IoU(%) $\uparrow$	Acc(%) $\uparrow$
U-Net (static teacher)	–	0	499	67.34	57.28	93.03	0	474	69.95	60.26	96.67
U-Net	MICCAI	12 (2.5%)	0	50.00	39.71	92.35	13 (2.5%)	0	43.52	31.03	93.52
U-Net	MICCAI	51 (10%)	0	65.04	54.77	94.76	54 (10%)	0	67.39	55.76	96.75
U-Net	MICCAI	103 (20%)	0	70.94	61.79	95.06	108 (20%)	0	67.95	58.22	96.93
U-Net	MICCAI	517 (100%)	0	81.68	73.74	96.65	542 (100%)	0	74.81	65.56	97.29

TABLE I: Comparison of segmentation performance on BUSI and UBB (UDIAT + BREASTUSG + BUSUCLM) datasets under different labeling ratios.

Method	APPG	UEWF	AURCL	Dice(%)	IoU(%)
SupOnly				50.00	39.71
I		✓		54.09	42.93
III	✓			71.17	61.63
IIII	✓		✓	71.64	62.49
IIIII	✓	✓		72.16	62.56
IIIIII	✓	✓	✓	72.72	63.11

TABLE II: Ablation study of different components on BUSI dataset under 2.5% partition protocol. “SupOnly” stands for training using only labeled data.

Fig. 5: Visual comparison of different ablation study on the BUSI and UBB datasets. All models are trained in 2.5% labeled data. Red, green and yellow regions represent ground truth, prediction and overlapping regions, respectively. (a) image, (b) ground truth, (c) APPG (step1), (d) STWP (step 2), (e) Ours

Method	Dice (%)	IoU (%)
w/o Patchify	71.89	62.52
w/ Patchify	72.72	63.11

TABLE III: Ablation study on the effectiveness of the patchify in UEWF under 2.5% partition protocol.

$\tau$	Dice (%)	IoU (%)
60	71.81	62.35
40	71.96	62.45
20	72.72	63.11

TABLE IV: Ablation study on the confidence threshold

\tau

in the AURCL module under 2.5% partition protocol.

Fig. 6: Visual comparison of zero-shot bounding boxes and pseudo masks under different prompt conditions. (a) Bounding boxes generated using the medical term “tumor” ; (b) bounding boxes generated using the medical term “lesion” ; (c) bounding boxes generated using the medical term “breast cancer” ; (d) bounding boxes generated using the radiology term “high density” ; (e) bounding boxes produced by our proposed “Appearance-prompted training-free pseudo label generation” strategy; Overlap denotes pseudo masks generated by applying SAM to the bounding boxes from our strategy, and the overlapping regions with GT.

Method	Venue	BUSI		UBB
Method	Venue	Dice(%)	IoU(%)	Dice(%)	IoU(%)
MediClipV2	MedIA’25	28.74	22.08	23.87	18.55
UniversalSeg	ICCV’23	30.68	21.99	48.17	35.77
GroundingDino(SFT)	ECCV’24	44.14	33.09	49.67	36.40
Ours	–	72.72	63.11	75.75	65.86

TABLE V: Comparison with VLM-based few-shot methods on BUSI and UBB under the 2.5% labeled partition protocol. All methods use 2.5% ground-truth labels for few-shot setting.

Fig. 7: Qualitative results on multiple datasets and imaging modalities to demonstrate the generalizability of our APPG pipeline. (a) skin lesion (dermoscopy), (b) thyroid nodule/tumor (ultrasound), (c) colorectal polyp (endoscopy), and (d) ovarian lesion (ultrasound).

Ultrasound Image Segmentation

Ultrasound image segmentation delineates organs, lesions, and other structures in ultrasound scans for measurement, diagnosis, and image-guided care.

See all 73 papers

Multiscale Switch for Semi-Supervised and Contrastive Learning in Medical Ultrasound Image Segmentation
Mar 19, 2026Jingguo Qu, Xinyang Han, Yao Pu +8Semi-Supervised Medical Image Segmentation Ultrasound Image Segmentation
Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hong Kong, China · Centre for Smart Health and School of Nursing, The Hong Kong Polytechnic University, Hong Kong, China · Department of Imaging and Interventional Radiology, The Chinese University of Hong Kong, Hong Kong, China
SSS: Semi-Supervised SAM-2 with Efficient Prompting for Medical Imaging Segmentation
Jun 10, 2025Hongjie Zhu, Xiwei Liu, Rundong Xue +5Promptable Image Segmentation Semi-Supervised Medical Image Segmentation
SWUN · MBZUAI · XJTU +2
Hybrid Diffusion Model for Breast Ultrasound Image Augmentation
Mar 27, 2026Farhan Fuad Abir, Sanjeda Sara Jennifer, Niloofar Yousefi +1Data Augmentation Synthetic Data Augmentation
Electrical and Computer Engineering, University of Central Florida · Department of Medicine, University of Central Florida College of Medicine · Department of Computer Science, University of Central Florida +1

Mar 19, 2026J/K move · Enter open · S save

Multiscale Switch for Semi-Supervised and Contrastive Learning in Medical Ultrasound Image Segmentation

Jingguo Qu, Xinyang Han, Yao Pu, Man-Lik Chui+7

Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hong Kong, China · Centre for Smart Health and School of Nursing, The Hong Kong Polytechnic University, Hong Kong, China · Department of Imaging and Interventional Radiology, The Chinese University of Hong Kong, Hong Kong, China

Semi-Supervised Medical Image Segmentation Ultrasound Image Segmentation Contrastive Learning Data Augmentation

Open paper AskPDF arXiv

Medical ultrasound image segmentation faces significant challenges due to limited labeled data and characteristic imaging artifacts including speckle noise and low-contrast boundaries. While semi-supervised learning (SSL) approaches have emerged to address data scarcity, existing methods suffer from suboptimal unlabeled data utilization and lack robust feature representation mechanisms. In this paper, we propose Switch, a novel SSL framework with two key innovations: (1) Multiscale Switch (MSS) strategy that employs hierarchical patch mixing to achieve uniform spatial coverage; (2) Frequency Domain Switch (FDS) with contrastive learning that performs amplitude switching in Fourier space for robust feature representations. Our framework integrates these components within a teacher-student architecture to effectively leverage both labeled and unlabeled data. Comprehensive evaluation across six diverse ultrasound datasets (lymph nodes, breast lesions, thyroid nodules, and prostate) demonstrates consistent superiority over state-of-the-art methods. At 5% labeling ratio, Switch achieves remarkable improvements: 80.04% Dice on LN-INT, 85.52% Dice on DDTI, and 83.48% Dice on Prostate datasets, with our semi-supervised approach even exceeding fully supervised baselines. The method maintains parameter efficiency (1.8M parameters) while delivering superior performance, validating its effectiveness for resource-constrained medical imaging applications. The source code is publicly available at https://github.com/jinggqu/Switch

Abstract

Ultrasound Image Segmentation

Explore similar work

Multiscale Switch for Semi-Supervised and Contrastive Learning in Medical Ultrasound Image Segmentation

SSS: Semi-Supervised SAM-2 with Efficient Prompting for Medical Imaging Segmentation

Hybrid Diffusion Model for Breast Ultrasound Image Augmentation