The incidence of breast cancer has been increasing year by year, with its global incidence second only to lung cancer. It is the most common malignant tumor in women and the leading cause of cancer-related death (1). Early detection, timely diagnosis, and prompt treatment are crucial for improved survival and decreased mortality of breast cancer.

Non-mass breast lesions (NMLs) are discrete abnormalities that distinctly differ from normal breast tissue, lack the margination of a mass and cannot be assigned a specific shape. These lesions are often subtle and may visualized primarily in one plane only (2). NMLs accounted for 9.21% of all breast lesions (3). The pathological types of NMLs are diverse. The most common breast cancers identified as non-mass findings on US images were ductal carcinoma in situ (DCIS) or invasive lobular carcinoma (ILC) (4). Benign NMLs include fibrocystic change, sclerosing adenosis, radial scar, granulomatous mastitis, and intraductal papilloma, etc (5).

Ultrasound (US) is one of the most commonly used imaging methods for breast cancer screening. The American College of Radiology Breast Imaging-Reporting and Data System (ACR BI-RADS) ^® Atlas (6) has described US features of mass-like breast lesions in a standardized format clearly and concisely. However, there’re some limitations when it’s applied to NMLs. Boundaries and margins can be used to determine whether a mass is benign or malignant, while the boundaries of NMLs are often blurred. For example, DCIS is distributed along the duct and the extent of the lesion is difficult to define accurately (7). In such cases, the assessment of the margins is almost ineffective.

Mammography is highly sensitive in detecting microcalcifications. However, it involves radiation exposure and shows relatively low sensitivity for diagnosing dense breast tissue (8). Magnetic resonance imaging (MRI) is widely recognized for its high sensitivity in the diagnosis of breast cancer. However, its clinical utility is often constrained by factors such as high cost, prolonged examination time, and limited sensitivity in detecting microcalcifications. CEUS provides real-time dynamic visualization of microcirculatory perfusion, demonstrating hemodynamic assessment capabilities comparable to MRI (9), while offering advantages in accessibility, shorter examination duration and patient convenience. CEUS is effective in differentiating benign from malignant mass-like breast lesions. However, evidence regarding its application in NMLs remains limited. First, standardized quantitative criteria for CEUS assessment of NMLs are lacking. Second, the value of CEUS in optimizing BI-RADS category for NMLs and its impact on reducing unnecessary biopsies has not been sufficiently investigated. Accordingly, this study aims to systematically investigate the diagnostic value of a CEUS-optimized BI-RADS category prediction model for NMLs and its impact on biopsy decision-making. The specific objectives are as follows: (1) to identify CEUS characteristics for distinguishing benign and malignant NMLs; (2) to evaluate the diagnostic performance of the CEUS prediction model in differentiating NMLs; (3) to explore the role of CEUS in optimizing BI-RADS category of NMLs, thereby providing evidence for reducing misdiagnoses and unnecessary biopsies.

The conventional US features of NMLs are characterized by diverse morphologies, indistinct boundaries, and the absence of typical mass-occupying effects. These attributes reduce the diagnostic accuracy of the BI-RADS category for NMLs compared with mass-like lesions, rendering them a major challenge in conventional US diagnosis. According to the findings of Ko et al., all NMLs were classified as BI-RADS 4A or above based on their grayscale US features. If biopsies were performed on all of them, it would lead to considerable overuse of medical resources. This study aimed to construct a CEUS prediction model to optimize the BI-RADS category of NMLs and reduce unnecessary biopsies.

Previous studies (16, 17) have demonstrated that microcalcification is an independent risk factor for malignant NMLs. Drawing on the findings of Ko et al., NMLs with microcalcifications (type Ib and IIb) were classified as BI-RADS 4C in this study. These lesions exhibited a malignancy rate as high as 93.55%, significantly higher than lesions without microcalcifications. Therefore, we recommended that all NMLs with microcalcifications should be biopsied to confirm the pathological diagnosis. Conversely, the malignancy risk of NMLs without microcalcifications varied substantially. Given that conventional US alone was insufficient for accurate differentiation, the incorporation of CEUS was necessary to improve diagnostic accuracy. The CEUS prediction model established in this study provided excellent diagnostic performance for differentiating benign and malignant NMLs, with an area under the ROC curve (AUC) of 0.876 and an accuracy rate of 87.8%.

Univariate analysis of CEUS qualitative features in this study demonstrated that earlier wash-in, later wash-out, hyperenhancement, heterogeneous enhancement, enlarged enhancement extent, and crab claw-like enhancement were the main indicators of malignant NMLs. These were consistent with previous research findings (18, 19). Angiogenesis is a vital process for genesis and progression of the tumor, preceding the morphological alterations. Breast cancer cells secrete vascular endothelial growth factor (VEGF), which induces extensive neovascularization (20–22). The neovascular walls are structurally immature and highly permeable. Tumor blood vessels exhibit disorganized and tortuous architecture, with widened endothelial gaps. Consequently, the CEUS features of breast cancer were earlier wash-in, later wash-out, and hyperenhancement. There are heterogeneous regions such as necrosis, fibrosis, and hemorrhage within the tumor and disparate spatial distribution of blood vessels, resulting in heterogeneous enhancement on CEUS (23). Binary logistic regression analysis revealed that enlarged enhancement extent and crab claw-like enhancement were independent risk factors for malignant NMLs. Although both benign and malignant NMLs display indistinct boundaries and irregular shapes on US, malignant tumors grow in an invasive manner, with the peritumoral tissue hyperplasia area containing new blood vessels. The new blood vessels of malignant tumors often arrange in a radial pattern centered on the tumor (13, 24). Therefore, enlarged enhancement extent and crab claw-like enhancement can be observed on CEUS. In contrast, most benign NMLs exhibit non-invasive growth patterns, without abnormal neovascularization. The majority of benign NMLs don’t display the typical CEUS features associated with malignant lesions. However, a minority of benign NMLs demonstrate CEUS features similar to malignancy, due to imbalanced angiogenesis regulation or potential malignant biological behaviors. In this study, 13 benign NMLs were misdiagnosed as malignant after CEUS optimization, including atypical hyperplasia, intraductal papilloma, inflammation, sclerosing adenosis, and fibrocystic mastopathy. The underlying reason is that atypical hyperplasia and intraductal papilloma are precancerous lesions with active cellular proliferation and increased microvessel density. These features result in hyperenhancement on CEUS, which may lead to misdiagnosis (25–27). Inflammatory stimulation causes local vasodilation and increased blood flow within breast lesions. As the disease progresses, inflammatory cells infiltrate the surrounding tissue (28, 29), which may manifest as malignant CEUS features. Although sclerosing adenosis is a benign condition, it is often associated with capillary proliferation within the lobules and stroma, as well as stromal sclerosis and glandular distortion. These histopathological features can create a “pseudo-invasive” appearance, which may lead to misdiagnosis as malignancy (30). Fibrocystic mastopathy may present with malignant features on imaging, such as punctate hyperechoic foci, which are often attributable to micro-echo enhancement of the cyst wall and the posterior region. These findings can be easily misinterpreted as calcifications. In such cases, the local magnification function should be utilized, to carefully observe the internal structure of the mass and minimize the risk of misdiagnosis (31). In the CEUS prediction model, 5 malignant NMLs were misdiagnosed as benign, including DCIS and IDC with mucinous carcinoma. Ductal carcinoma in situ (DCIS) is characterized by the confinement of tumor cells within the mammary ducts, without invasion beyond the basement membrane. The ability of these tumor cells to stimulate the surrounding stromal vasculature is relatively limited. DCIS typically demonstrates low expression levels of VEGF, which contributes to a significantly reduced microvessel density (32). Moreover, CEUS is insensitive to microcalcifications, which represent a characteristic imaging feature of DCIS. This limitation may result in the misdiagnosis of certain DCIS lesions on contrast-enhanced imaging. Mucinous carcinoma is particularly prone to misdiagnosis on CEUS because it may exhibit benign features, such as expansive growth and relatively poor vascularity. This tendency is primarily attributable to the abundant extracellular mucin, pseudocapsule formation, and sparse blood vessels (33, 34).

To address the aforementioned issues of misdiagnosis, a multimodal imaging assessment can compensate for the limitations inherent to any single diagnostic modality. The diagnostic efficacy of mammography is significantly reduced in dense breasts, and the prevalence of breast density among Asian women is much higher than that in Western populations. A proportion of patients present primarily with localized breast symptoms, such as palpable masses or nipple discharge, and mammography is less accurate than ultrasound in differentiating between cystic and solid breast lesions. Mammography requires patients to be exposed to ionizing radiation, and breast compression is necessary during image acquisition, which is poorly tolerated by some patients due to discomfort. These factors have limited the application of mammography in this study. However, mammography demonstrates substantially higher sensitivity than CEUS for the detection of microcalcifications and is therefore more effective in identifying malignant lesions, such as DCIS. In addition, imaging findings should be interpreted in conjunction with clinical information and individual risk factors. For patients with high-risk characteristics, including advanced age or a family history of breast malignancy, lesions that appear benign on CEUS but suspicious on other imaging modalities should prompt a recommendation for biopsy. Conversely, the possibility of inflammatory lesions causing false-positive imaging findings should be carefully considered in patients presenting with localized pain, a history of trauma, or signs of inflammation.

If biopsy was performed on NMLs categorized as BI-RADS 4A or above based on conventional US, the biopsy rate reached 100% in this study. However, the malignancy detection rate was only 52.03%. It indicated that US-BI-RADS tended to overdiagnose a large number of benign NMLs, leading to unnecessary biopsies or surgical interventions. The BI-RADS categories of 148 NMLs were optimized by a CEUS prediction model in this study. Setting CEUS-optimized BI-RADS 4A as the biopsy threshold, the biopsy rate decreased to 75.00% while the malignancy detection rate increased to 66.67%. This approach avoided unnecessary biopsy for 37 lesions classified as CEUS-optimized BI-RADS 3, thereby reducing interventions for benign lesions. The missed diagnosis rate was only 3.90%. If CEUS-optimized BI-RADS category was used to guide breast biopsy, the detection rate of breast cancer increased and the malignant risk remained at a relatively low level. This provided valuable reference for making rational diagnosis and treatment decisions in clinical practice.

This study has some limitations. First, as a retrospective analysis, selection and recall biases were unavoidable. Second, sample capacity was small and diversity of pathology was relatively limited in this study. Therefore, multicenter studies with larger cohorts were warranted for validation. Third, only CEUS was used as the sole imaging modality, without integration of multimodal imaging approaches, such as elastography, mammography, or contrast-enhanced magnetic resonance imaging (MRI), which may have resulted in an incomplete assessment of lesion characteristics. In addition, CEUS itself has inherent limitations. Its interpretation relies heavily on the operator’s clinical experience and subjective judgment, and interobserver variability may compromise diagnostic consistency. Moreover, there is currently no universally accepted and standardized system for interpreting CEUS enhancement patterns; the evaluation criteria applied vary across studies, limiting the comparability of results. Finally, CEUS has limited sensitivity for detecting microcalcifications, which are often critical diagnostic clues for certain malignant lesions. This shortcoming may adversely affect early lesion detection and diagnostic accuracy.

The CEUS prediction model offered an optimized approach to the BI-RADS category of NMLs. Implementing CEUS-optimized BI-RADS category to guide core needle biopsy may not only minimize unnecessary procedures but also improve the detection rate of malignant breast lesions.