Granulomatous lobular mastitis (GLM) is a non-infectious, nonspecific inflammatory breast disease. In recent years, its rising incidence and increasingly complex clinical manifestations, coupled with a lack of robust evidence-based medical data, have rendered GLM a formidable diagnostic and therapeutic challenge [1]. Accurate and timely diagnosis is essential for initiating effective treatment and improving clinical outcomes. Owing to its speed, convenience, non-invasiveness, and absence of ionizing radiation, breast ultrasonography has become the preferred imaging modality for the diagnosis of GLM [2]. However, in the past, ultrasound primarily focused on detecting breast masses and their advanced stages. In particular, during the non-mass stage, limitations in probe frequency and resolution hinder the identification of subtle ductal changes and periductal abnormalities, contributing to delayed or missed diagnoses. Nonetheless, early detection remains critical, as it offers the greatest potential for effective intervention and screening.

Our institution is recognized as one of the earliest and most prominent centers for the diagnosis and treatment of GLM in China [3-5]. Over the past decade, our center has accumulated extensive clinical experience through a high volume of cases. This has enabled us to build a substantial institutional repository of ultrasound images and clinical data, forming a solid foundation for research. Through a systematic analysis of this repository, we previously identified and described a set of stage-specific ultrasonographic features of GLM. Building upon this foundation, the present study aims to further evaluate and validate the reliability of these ultrasound features, providing objective statistical evidence to support their diagnostic value. The primary aim is to comprehensively delineate the ultrasonographic manifestations of GLM across its various clinical stages—with a particular focus on the diagnostically challenging non-mass stage—to provide a valuable reference for clinical practice.

All cases in the validation cohort were derived from female patients aged 18 years and older, who were diagnosed with GLM either by initial assessment at external hospitals or through outpatient consultation at Longhua Hospital, Shanghai University of Traditional Chinese Medicine, between June 2022 and December 2023. A total of 273 patients were enrolled. The study, which built upon prior retrospective analysis of historical ultrasound images for feature summarization, was conducted under the approval of the Ethics Committee of Longhua Hospital, Shanghai University of Traditional Chinese Medicine (approval number: 2021LCSY047), and all patients in the validation cohort provided informed consent.

Based on the clinical and imaging characteristics of GLM, disease progression was classified into 6 stages: Non-Mass Stage, Mass Stage, Abscess Stage, Late Ulceration Stage, Fistula Stage, and Complex Type (defined as the simultaneous presence of ≥ 2 ultrasonographic features corresponding to different GLM stages within the same breast, characterized by the coexistence of Mass, Abscess, Late Ulceration, and Fistula stages). In our preliminary research, we summarized a set of stage-specific ultrasonographic features of GLM, including five distinct types of the “track sign,” the “donut sign,” the “false burr sign,” features indicative of abscess formation, post-ulceration manifestations, and duct-like hypoechoic patterns, among others.

Clinical assessment of each case was determined through multidisciplinary evaluation involving two senior breast specialists, one radiologist, and one pathologist. This assessment was based on findings from specialized clinical examinations, MRI, and pathological biopsy. Ultrasound assessment was independently performed by two senior sonographers, who evaluated each case using the predefined ultrasonographic criteria developed during our preliminary studies. In cases where the assessments differed, a consensus meeting was held to reach a final diagnosis. The sonographers were blinded to the patients’ clinical and pathological results during image interpretation to minimize bias.

Ultrasonic examinations were performed using the Hitachi ALOKA ARIETTA 850 color Doppler ultrasound system, equipped with the L64 probe operating within a probe frequency range of 5-18 MHz. Patients were positioned in the supine position with both arms relaxed and placed at their sides. The patient's position could be adjusted during the examination as needed. Scanning was performed in a radial pattern from the nipple, utilizing fan-shaped and circular multi-sectional scanning techniques. The following parameters were recorded: lesion location, size, shape, echogenicity, blood flow signals, duct diameter, presence of thickening around the duct, secretions within the duct lumen, and calcification.

Examination flow and operational details:

① A routine ultrasound (US) examination of the lesion was conducted, documenting the lesion's location, size, shape, echogenicity, blood flow signals, duct diameter, and calcification.

② The acoustic beam was focused on the region of interest.

③ The “ZOOM” function was applied to magnify and visualize the entire duct in cases of ductal dilation, observing for lipid secretions, whether the duct extends toward the nipple, and assessing the smoothness or roughness of the duct wall, as well as any thickening around the duct and reduction in interstitial echogenicity.

④ A coupling agent was applied generously to the areola, and gentle pressure was applied to the probe without excessive force, preventing ductal closure.

⑤ The color scale was adjusted to the lowest feasible setting, ideally below 2 cm/s, and blood flow signals within the hypoechoic area of the duct were carefully examined.

⑥ For suspected lipid secretions, light probe compression was applied, allowing the visualization of low echogenicity or fine, punctate high echogenicity patterns within the duct, with a floating or flowing sensation.

The clinical and ultrasound assessment results are summarized in table 1. A total of 261 cases were clinically assessed as GLM, while 12 cases were not GLM, including 4 cases of breast cancer, 2 cases of intraductal papilloma, and 4 cases of breast hyperplasia. Ultrasound assessment of GLM demonstrated a sensitivity of 98.85%, with high specificity of 83.33% and positive predictive value of 99.23%. To further evaluate diagnostic reliability, we calculated the positive likelihood ratio (LR⁺ = 5.93; 95% CI: 1.82–19.29), which—together with the high sensitivity—supports the reliability and stability of the ultrasonographic diagnostic performance for GLM.

The distribution of ultrasonographic features across different stages of GLM in 261 patients is presented in table 2. Cases categorized as Complex Type were excluded from the main statistical analysis because their overlapping ultrasonographic features across multiple stages could confound comparisons among mutually exclusive groups. Therefore, the Chi-square tests in table 2 were conducted only for cases assigned to a single, clearly defined stage to avoid confounding the interpretation of stage-specific ultrasonographic findings. We observed a significant association between specific ultrasonographic features features and the clinical stage of GLM. Various types of track signs, with the exception of the track sign with periductal abscess formation, were significantly more prevalent in the Non-Mass Stage compared to other stages. Notably, the track sign with periductal abscess formation also demonstrated a higher tendency to appear in the Non-Mass Stage. The donut sign was significantly more frequent in both the Non-Mass Stage and Abscess Stage than in other stages. The false burr sign was predominantly observed during the Mass Stage, showing a significantly higher distribution compared to other stages.

Another characteristic ultrasonographic finding of the Abscess Stage was the presence of large areas of mixed echogenicity with irregularly thickened and rough abscess walls; a flowing sensation within the abscess cavity was elicited upon probe compression. In the Late Ulceration Stage, a specific ultrasonographic feature was the appearance of a relatively well-defined and smooth-edged hypoechoic area in the areolar region, which communicated with the ductal system. In the Fistula Stage, the characteristic feature was a duct-like hypoechoic structure extending toward the skin, with evident thickening of the inner ductal wall.

Building on these validated ultrasonographic features, we provide a comprehensive overview of the characteristic and stage-specific imaging manifestations of GLM, supplemented with representative figures to enhance clinical recognition and interpretation.

In the early stages of GLM, the disease may present with pain as the only clinical manifestation, often leading to misdiagnosis as mammary gland hyperplasia (a disorder of breast structure), corresponding to the Non-Mass Stage. The underlying pathological basis at this stage involves irregular hyperplasia of ductal epithelium, ductal dilation, and periductal tissue fibrosis. As a result, the ultrasonographic appearance exhibits similar characteristics, with dilated and thickened ductal cross-sections resembling a "train track" structure, known as the track sign. As the extent of inflammatory cell infiltration increases and the severity of the disease progresses, the sonographic features evolve. Five distinct types of ultrasonographic changes have been summarized, although these five types of track signs do not necessarily appear in isolation; different track signs may coexist in the same patient.

The ultrasonographic appearance shows thickening of the ductal wall, with the inner wall displaying a strip-like or linear high echogenicity. The ductal wall is smooth, and lipid-containing secretions are seen floating within the duct. A thin, uniformly hypoechoic halo is observed around the duct (Fig. 1).

The ultrasonographic appearance shows a single or several ducts behind the areola with irregular thickening of the ductal walls. The surrounding ductal wall is hypoechoic, and the duct is notably dilated. A large amount of hyperechoic secretion is accumulated within the lumen. Upon probe compression, a floating sign is observed (Fig. 2).

The ultrasonographic appearance shows significant thickening of the inner wall of some ducts, presenting as isoechoic. The surrounding ductal wall is further thickened and hypoechoic. A small amount of anechoic secretion is accumulated within the duct (Fig. 3).

As the disease progresses, the ultrasonographic appearance shows high echogenicity on the inner wall of the duct, with further thickening of the surrounding ductal wall, presenting as a hypoechoic area that merges with the surrounding stroma, forming a large hypoechoic region. A small amount of pus is observed accumulating around the duct. Clinically, some cases present with palpable localized thickening and flaky nodules, signaling the transition from the non-mass stage to the mass stage (Fig. 4).

This represents the most severe lesion manifestation in the non-mass stage, marking the transition from the non-mass stage to the mass stage. At this point, the inflammatory response within the duct is at its peak, and the duct diameter is widest. Contrast-enhanced ultrasound at this stage reveals ring-like enhancement of the dilated ductal walls, which provides valuable information about the location and extent of the lesion (Fig. 5).

The ultrasonographic features observed at this stage warrant particular attention, as the mass-like lesions must be differentiated from breast carcinoma. Ultrasound typically reveals an ill-defined hypoechoic mass with angular margins and the appearance of a false burr sign, which may be associated with compressive expansion of the lesion into adjacent tissue planes and its expansive growth pattern [6]. The pathological basis of the false burr sign lies in the adhesion of mammary trabeculae or the formation of fibrous strands. Color Doppler imaging often shows relatively abundant internal vascularity within the mass. In addition, localized skin thickening and edema may be present, and in some cases, thickening of Cooper’s ligaments can also be observed (Fig. 6).

Patients with GLM may develop abscesses. On ultrasonography, this is manifested as large areas of mixed echogenicity in the areolar or peripheral regions. The inner wall of the abscess cavity appears irregularly thickened, with overlying skin edema and thickening. A substantial amount of purulent fluid is present within the cavity, and a flowing sensation can be elicited upon probe compression. Dilated ducts extending toward the nipple can also be observed [7] (Fig. 7).

Following the abscess stage, the disease progresses to the late ulceration stage, during which the extent of the lesion is significantly reduced, and the inflammatory response associated with ductitis and periductal inflammation is markedly alleviated. On ultrasound, a hypoechoic or mixed echogenic area with relatively smooth and well-defined margins can be observed in the periareolar region. This area communicates with the ductal system and extends toward the nipple (Fig. 8).

In some patients with GLM, the disease may progress to the fistula stage [8]. Ultrasonographically, this stage is characterized by the presence of one or more quadrants showing irregular sheet-like or duct-like hypoechoic structures traversing from the glandular layer through the adipose tissue and extending into the subcutaneous region. These structures often interconnect with one another and, in some cases, extend as deep as the retromammary space. The inner walls of the ducts are thickened, with or without associated pus accumulation. Skin ulceration may also be observed (Fig. 9).

In patients with GLM whose disease duration exceeds one month, it is common to observe a combination of clinical features such as masses, abscesses, ulcers, and fistulas. In other words, the progression of GLM in some patients does not follow a linear sequence from the non-mass stage to the mass stage, abscess stage, late ulceration stage, and fistula stage. Instead, the disease may follow a fluctuating course, with some lesions stagnating at a particular stage while new lesions emerge or progress into subsequent stages. This results in a complex pattern of overlapping old and new lesions (Fig. 10).

Ultrasound is the first-line imaging modality for GLM, aiding in the detection of inflammatory lesions, abscess formation, and fistulous tracts. It also facilitates ultrasound-guided biopsy and serves as a valuable tool for monitoring disease remission or progression during follow-up [9]. Although the imaging features of GLM are generally considered nonspecific, complex, and variable, a combination of clinical context and stage-specific ultrasonographic characteristics can significantly improve diagnostic accuracy. Common ultrasonographic findings include a hypoechoic or heterogeneous mass, an irregular hypoechoic lesion with tubular extension [10], and abscess formation. The diagnostic accuracy of ultrasound is closely related to the ultrasonographic subtype of GLM [11], with higher accuracy observed in patchy heterogeneous echo types and nodular or mass-forming types, while diffuse-type GLM remains more challenging to diagnose.

Early-stage GLM is the most easily overlooked phase in clinical practice and is frequently misdiagnosed as mammary ductal hyperplasia. In both domestic and international literature, the ultrasonographic features of GLM in the non-mass stage have been briefly summarized as parenchymal heterogeneity and distortion with or without acoustic shadowing [12], without further detailed characterization. Based on case review and validation at our center, we highlight that the presence of dilated and enlarged ductal cross-sections—collectively referred to as the track sign—is a key diagnostic indicator during this stage. Furthermore, we propose a five-type classification system for the non-mass stage based on disease severity. Among them, the latter two types reflect the most extensive inflammatory cell infiltration, characterized by significantly widened ductal lumens and more pronounced periductal thickening. These may appear as large, irregular or nodular hypoechoic structures within the gland. Importantly, these nodular structures are not true solid granulomas but are formed by the confluence of dilated ducts and surrounding fibrotic tissue [13,14].

Histologically, these findings are accompanied by focal fat necrosis, periductal fibrosis, and lymphocytic infiltration. On physical examination, broad regions of indurated nodularity can often be palpated. It is noteworthy that these five ultrasonographic patterns do not occur in isolation, nor do they necessarily follow a sequential progression. Multiple ultrasonographic features may coexist across different stages of the disease, in various regions of the same breast, or even bilaterally. Therefore, it is essential to recognize the characteristic track sign and distinguish it from other types of intraductal lesions. We believe that early detection of GLM should be a primary focus in clinical management. Enhancing awareness and understanding of early-stage GLM may help improve diagnostic accuracy.

In addition to detailing the ultrasonographic features of the non-mass stage, this study also summarizes the imaging characteristics of GLM across its full spectrum of clinical stages—including the mass stage, abscess stage, late ulceration stage, fistula stage, and complex type. Notably, complex-type GLM presents with diverse and overlapping ultrasonographic findings. During ultrasound examination, the sonographer must carefully document the condition of each lesion, including the location and extent of masses and abscesses, whether pus cavities are interconnected, the depth of any ulcerated tract from the skin surface, and the number, position, and orientation of fistulas. These ultrasound findings can guide clinicians in determining the appropriate timing and method of surgical intervention.

Moreover, we have standardized a systematic ultrasound examination protocol and operational procedure specific to GLM. Considering the common overlap, inconsistency, and ambiguity in ultrasonographic terminology for breast imaging, this study provides clearly defined sonographic terms and representative images to improve diagnostic clarity. This effort not only enhances clinicians’ understanding of GLM-specific ultrasound features but also lays the groundwork for more standardized reporting and examination practices in breast ultrasonography.

One limitation of this study is the lack of quantitative interobserver agreement analysis (e.g., Kappa statistics). The ultrasonographic assessments were conducted independently by two senior sonographers, and discrepancies were resolved through consensus discussion. As this study primarily focused on the qualitative characterization and validation of previously undefined imaging features, quantitative analysis such as interobserver Kappa testing was not included at this stage. Future research will incorporate Kappa consistency analysis and quantitative imaging evaluation methods to further enhance the objectivity and reproducibility of ultrasonographic assessment. Furthermore, this study did not include a comparative analysis with magnetic resonance imaging (MRI). With the growing use of MRI in GLM assessment, we acknowledge its valuable complementary role in delineating disease extent and aiding in staging [15]. Nonetheless, ultrasound remains the first-line and most accessible imaging modality for GLM, offering real-time, dynamic, and radiation-free assessment that cannot be replaced in routine clinical practice. In future research, we plan to perform a systematic comparison between ultrasonographic and MRI features across various GLM stages to better clarify their respective diagnostic strengths and complementary value in disease staging.

In conclusion, the findings of this study demonstrate that specific ultrasonographic features of GLM can be reliably used to identify the disease and determine its clinical stage. We established a standardized operating procedure for ultrasonographic evaluation of GLM, systematically characterized the ultrasonographic features across different stages, and provided representative imaging examples. These efforts aim to improve the accuracy of ultrasound-based diagnosis and staging of GLM, thereby enhancing the reproducibility and applicability of this diagnostic approach in clinical practice.