A novel qualitative pattern of shear wave elastography for differentiating suspicious metastatic axillary lymph nodes on B-mode ultrasound: a multi-center retrospective study.

Introduction
Breast cancer is the most prevalent malignant tumor among women in world. Axillary lymph node (ALN) metastasis is the most significant predictor of overall recurrence and survival in patients with breast cancer. Accurate assessment of ALN status is a critical component of breast cancer staging and clinical decision-making (1). Thus, it is essential that a reliable, noninvasive method for the preoperative evaluation of ALN status be developed.
The most commonly used method for the preoperative evaluation of ALN status in clinic is conventional ultrasound, which mainly focuses on morphological characteristics, including a longitudinal-to-transverse ratio (L/T) <2 and the absence of hilar and cortical thickening (2). This approach is convenient, cheap, noninvasive, and safe while lacking radiation exposure to patients. However, due to an overlap of morphological characteristics, conventional ultrasound has low sensitivity in differentiating between benign and metastatic ALNs (3).
Shear wave elastography (SWE) can quantitatively and qualitatively evaluate tissue stiffness and has been widely used in the differentiation of benign and malignant breast lesions (4,5) and the prediction of metastatic ALNs (3,6). No findings, vertical stripes, and spot patterns on breast lesions are critical indicators of negative ALN status (3). Conversely, stiff rim sign on breast lesions is highly indicative of positive ALN status (6). However, few studies have reported on the value of SWE in the differentiation of benign and metastatic ALNs. One study (7) found that qualitative SWE could improve accuracy in the diagnosis of metastatic ALNs in patients with breast cancer. In our clinical experience and preliminary test results, we found peritumoral stiffness in the color elastic map to be a typical sign of metastatic ALNs and localized green colored spots at the margin of ALNs to indicate benign ALNs. Thus, we developed a novel qualitative pattern classification system to discriminate metastatic from benign ALNs and improve the clinical use of SWE.
For ALNs exhibiting suspicious metastatic characteristics on B-mode ultrasound (BMUS), including an L/T <2 and the absence of hilar or cortical thickening >3 mm, SWE may provide additional stiffness features to improve diagnostic differentiation. We conducted a study aimed at validating a novel qualitative SWE pattern classification system for discriminating metastatic and benign ALNs in BMUS-suspicious cases, potentially offering a noninvasive and precise diagnostic alternative. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-2008/rc).

Methods
This retrospective study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments, and was approved by the Ethics Committee of the Yueyang Central Hospital (approval No. 2024-005). Another two participating hospitals were also informed of and agreed to the study. The ethics committees waived the requirement for written informed consent for participation due to the retrospective nature of the analysis.

Patients
A total of 144 patients with 144 ALNs attending Yueyang Central Hospital, Hunan Cancer Hospital/The Affiliated Cancer Hospital of Xiangya School of Medicine, or Linxiang People’s Hospital from May 2022 to June 2025 were enrolled. The inclusion criteria were as follows: (I) breast cancer confirmed by surgery, with all ALNs being confirmed via lymph node biopsy and axillary lymph node dissection (ALND); (II) SWE examinations performed within 2 weeks before lymph node biopsy or surgery; and (III) all ALNs exhibiting suspicious metastatic characteristics on BMUS, including an L/T <2 and the absence of hilar or cortical thickening >3 mm. Meanwhile, the exclusion criteria were as follows: (I) patients that were pregnant or lactating, (II) radiotherapy or chemotherapy administered prior to the ultrasound examination, (III) unsatisfactory SWE images of ALNs, and (IV) ALN biopsy performed before ultrasound examination.

Ultrasound examinations and imaging analysis
All BMUS examinations, color Doppler, and SWE examinations of ALNs from three hospitals were performed with Aixplorer ultrasound system (SuperSonic Imaging, Aix-en-Provence, France). All radiologists performing SWE examinations had undergone standardized SWE operation training. All imaging data from three hospitals were assessed by two radiologists, each with 10 years of experience in breast ultrasound and 5 years in SWE. The radiologists were blinded to each other’s findings and to the final pathological results. In cases of disagreement, a third, more experienced radiologist was consulted to reach a consensus.
BMUS was conducted on patients in a supine position, with their arms positioned above their heads. For ALNs, ultrasound images were obtained in a typical plane. The BMUS assessment of ALNs focused on the long axis and short axis, the presence of a hilum, and cortical thickening, with a thickness of >3 mm indicating potential metastasis. Lymph node morphology, classified into types I–VI, was determined according to the relationship between the cortex and the hilum as follows (8): type I, ALN with an imperceptible cortex and preserved hilum; type II, ALN with a thin (≤3 mm) cortex and preserved hilum; type III, ALN with a diffuse hypoechoic cortex (>3 mm) and preserved hilum; type IV, ALN with a generalized lobulated cortex (>3 mm) and hilum; type V, ALN with focal and/or eccentric hypoechoic cortical lobulation and effacement and/or displacement of the hilum; and type VI, ALN with absent and/or replaced hilum appearing as a hypoechoic mass.
Additionally, the Angio-Plus (AP) microvascular Doppler ultrasound technique (SuperSonic Imaging) was used to evaluate the vascular distribution of ALNs. AP is an innovative microvascular Doppler ultrasound technique implemented on the Aixplorer platform (SuperSonic Imaging). Its technical foundation rests on two key pillars: the transmission of unfocused plane waves and advanced three-dimensional wall filtering. By emitting these unfocused waves at the maximum permissible pulse repetition frequency, the system can reconstruct all image pixels from a single insonification. This approach achieves a significantly higher sampling rate compared to classical color Doppler flow imaging. The vascular distribution of ALNs in AP was classified into four patterns as follows: vascular pattern 1, no obvious vascularity inside the ALNs; vascular pattern 2, portal blood flow signal inside the ALNs; vascular pattern 3, vessels at the margin of ALNs; and vascular pattern 4, combined vessels inside and at the margin of ALNs.
All SWE examinations were performed with a L10–5 linear array transducer under standardized protocols with minimal transducer pressure. The scanning protocol included both the target ALN and adjacent normal tissue within the sampling frame, with a standardized color-coded stiffness scale (0–180 kPa; blue = soft; red = stiff) being applied. Quantitative measurement was performed via a dual region-of-interest (ROI) approach with 2×2 mm2 Q-Box measurement tools. The first Q-Box was placed in the hardest part of the cortical zone of the ALN, and the minimal value (Emin), mean value (Emean), maximum value (Emax), and standard deviation (SD) were obtained automatically. The second Q-Box was positioned within the adjacent normal tissue for reference. The elasticity ratio (Eratio), calculated automatically by the ultrasound system, represented the quotient of cortex-to-normal tissue stiffness values.
We developed a novel qualitative color pattern SWE classification system in this study, in which the SWE images of ALNs are classified into five color patterns (Figure 1) as follows: color pattern 1, homogeneous blue at the margin and inside the ALN; color pattern 2, green spots visible at the margin of the ALN; color pattern 3, heterogeneous green, yellow, or red areas visible inside the ALN; color pattern 4, filling defect within the ALN, with the other region of the color map presenting homogeneous blue or green spots; and color pattern 5, a localized colored stiff rim or spots at the margin of ALN, with the color of the stiff rim or spots being mainly yellow and red, with or without a filling defect within the ALN.

Statistical analysis
Statistical analyses were conducted with SPSS version 23.0 (IBM Corp., Armonk, NY, USA) and MedCalc version 19.0 (MedCalc Software, Ostend, Belgium). Continuous variables were analyzed via independent t-tests, while categorical variables were assessed with Chi-squared tests. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and the area under the receiver operating characteristic curve (AUC) of BMUS characteristics, BMUS patterns, vascular distribution, and SWE parameters were calculated and compared. A P value less than 0.05 was deemed statistically significant.

Results

General characteristics of all ALNs
A total of 144 patients with breast cancer, comprising 144 ALNs, were enrolled in this study. ALND was performed in 52 (36.1%) patients, while 92 (63.9%) underwent sentinel lymph node biopsy (SLNB). Histopathological assessment identified metastatic involvement in 61 (42.4%) cases, with the other 83 (57.6%) patients exhibited benign findings. The clinical pathological characteristics of all breast lesions and ALNs are listed in Table 1.

Suspicious BMUS morphology and vascular distribution characteristics in differentiating between benign and metastatic ALNs
The BMUS and vascular distribution characteristics differentiating between benign and metastatic ALNs are listed in Table 2. An L/T of <2, a maximum diameter, absence of the hilum, cortical thickening >3 mm, or BMUS patterns of ALNs were significantly different between the negative and metastatic ALNs. Moreover, benign ALNs often demonstrated no obvious vascularity or portal blood flow signal within them, while most metastatic ALNs had vessels at their margin or combined vessels within them and at the margin.

Quantitative SWE parameters and novel qualitative SWE patterns between benign and metastatic ALNs
The novel qualitative SWE patterns and quantitative SWE parameters between benign and metastatic ALNs are listed in Table 3. The benign ALNs had patterns 1 and 2, while metastatic ALNs tended to have color patterns 3, 4, or 5. There were significant differences in qualitative SWE patterns between benign and metastatic ALNs (P<0.05). Metastatic ALNs demonstrated significantly elevated mean values of Emin, Emean, Emax, SD, and Eratio as compared to benign ALNs.

Comparison of diagnostic performance for differentiating between benign and metastatic ALNs among BMUS characteristics, vascular distribution, and SWE parameters
Receiver operating characteristic (ROC) curve analysis was performed to evaluate the diagnostic performance of each parameter (Figure 2). Using the optimal cutoff values derived from the ROC analysis, we calculated the sensitivity, specificity, PPV, NPV, AUC, and overall accuracy (Table 4). The optimal cutoff value of Emin, Emean, Emax, SD, and Eratio were 4.35 KPa, 17.30 KPa, 41.55 KPa, 1.95 KPa, and 2.25, respectively. Compared with BMUS characteristics, vascular distribution, and quantitative SWE parameters, the novel qualitative SWE patterns achieved the highest diagnostic performance, with an AUC of 0.972, a sensitivity of 96.72%, a specificity of 97.59%, and an accuracy of 97.22%.

Discussion
In this study, we found that our proposed qualitative SWE patterns, as compared with BMUS characteristics, vascular distribution, and quantitative SWE parameters, had the best diagnostic performance in identifying BMUS-suspicious ALNs. As a radiation-free and highly reproducible approach, these novel qualitative SWE patterns allow for the precise discrimination of metastatic ALN characteristics. This capability facilitates optimized surgical planning (e.g., ALND vs. SLNB) and may reduce overtreatment, thereby improving quality of life of the patients being treated for breast cancer.
Previous studies (2,9,10) have reported cortical thickening >3 mm to be the most reliable BMUS characteristic for predicting metastasis, with a sensitivity of 56.3–68.8% and a specificity of 64.0–86.7%. However, the cortical thickening thresholds reported across studies are inconsistent, which can be attributed to the variations in study populations. Zhu et al. (11) identified 3.5 mm as the optimal cortical thickness threshold for metastasis prediction in patients with early breast cancer, with a 76% sensitivity and an 83% specificity. Pulappadi et al. (12) found that a cortical thickness ≥6.7 mm had the best diagnostic performance among BMUS features, achieving an 89.5% sensitivity and a 72.7% specificity. In our study, we used cortical thickening >3 mm as a cutoff value, which yielded a sensitivity and specificity of 100% and 22.89%, respectively. The high sensitivity might be related to the focus on BMUS-suspicious ALNs in our study.
Lymph node size provides limited diagnostic value when used as an isolated parameter. The morphological transition to a rounded configuration, characterized by decreased L/T ratio, constitutes a significant indicator of potential metastatic infiltration (13). In our study, an L/T ratio cutoff <2 had a sensitivity and specificity for differentiating benign and metastatic ALNs of 63.93% and 57.83%, respectively. Ultrasound sensitivity is significantly diminished in subcentimeter lymph nodes (<1 cm), as these frequently demonstrate benign spherical morphology (L/T ratio <2) and physiological hilar fat loss, mimicking malignant characteristics (14).
Moreover, vascular distribution of ALNs might be helpful in evaluating ALN status in patients with breast cancer. Characteristic hilar vascularity with symmetrical branching patterns can be visualized on Doppler ultrasound in normal lymph nodes (15). Metastatic lymph nodes often demonstrate abnormal peripheral or penetrating vascular patterns on imaging, which is attributable to tumor-induced neoangiogenesis. In our study, benign ALNs generally exhibited no obvious vascularity inside the ALNs or portal blood flow signal inside the ALNs, while vessels at the margin of ALNs or combined vessels inside and at the margin of ALNs were commonly present in metastatic ALNs. However, when we used vascular distribution patterns 1, 2, 3, 4 as the cutoff, there was significant overlap between benign and metastatic ALNs, with a sensitivity of 50.82%, a specificity of 95.18%, and an accuracy of 76.39%. This might be due metastatic lymph nodes with disorganized vascular patterns being readily identifiable in advanced disease stages; however, the early detection of subtle vascular abnormalities remains diagnostically challenging, even with the application of the microvascular Doppler ultrasound technique (AP). Thus, the diagnostic performance of these BMUS characteristics and vascular distribution was insufficient for accurately evaluating BMUS-suspicious ALNs status before surgery.
Previous studies have reported that quantitative SWE parameters have the ability to differentiate between benign and metastatic ALNs. Luo et al. (7) reported the AUC value of Emean (Emean >26.90 kPa) to be the highest for the differentiation of benign and metastatic ALNs. Ng et al. (16) reported that only two quantitative parameters (Emax and SD) were significant for detecting ALNs metastases, with an Emax >15.2 kPa yielding an AUC of 0.612. However, the optimal quantitative SWE parameters for differentiating between benign and metastatic ALNs vary across different studies. This may be because the values of quantitative SWE parameters (Emax, Emean, Emin, SD, and Eratio) are dependent on the size and place of the ROI. Moreover, a portion of metastatic ALNs present filling defects in color elastic maps, while the other regions of color map may present homogeneously blue, causing the quantitative SWE parameters to fall below the optimal cutoff values (Figure 3). Thus, the quantitative SWE parameters may not have the ability to correctly evaluate ALN status.
Compared with quantitative SWE parameters, qualitative SWE patterns have been reported to have better diagnostic performance for differentiating between benign and metastatic ALNs. Previous study (7) has used four color patterns: Benign ALNs typically exhibit color pattern 1, appearing as a homogeneous color on the elastic map, whereas metastatic ALNs often display color patterns 2 to 4, characterized by filling defects or localized colored areas at the margin. These have all demonstrated good diagnostic performance, with AUCs ranging from 0.813 to 0.983. However, we found the four-color patterns might not cover the entire color map of ALNs in patients with breast cancer. This is because a portion of ALNs present with heterogeneously green, yellow, or red areas visible inside the ALN or with a localized colored stiff rim at their margin. Thus, we developed five color patterns based on our clinical experience, in which benign ALNs present as color patterns 1 and 2, and metastatic ALNs generally present as color patterns 3 to 5.
In the early stages of metastasis, cancer cells typically invade the periphery of ALNs via afferent lymphatic vessels. The increase in tumor cell density induces changes in local tissue elasticity, often visible as a focal colored area in the marginal sinus. Subsequently, the cells spread to the medullary sinus and eventually throughout the entire lymph node, causing a marked increase in nodal stiffness (Figure 4). Over time, cancer cells progressively dominate the lymph node, infiltrate its capsule, and adhere to the surrounding tissues, which is accompanied by hyperplasia of perinodal fibrous tissue, causing a localized colored stiff rim at the margin of the lymph node (Figure 5).
Filling defect may be present in color patterns 4 or 5. Color pattern 4 might be interpreted as growth of cancer cells improved heterogeneity, including angiogenesis and necrosis, blocked the generation and/or transmission of shear waves. Meanwhile, color pattern 5 might be interpreted as cancer cells nearly dominating the entire lymph node, causing an increase in the stiffness of the lymph node that exceeds the threshold value of SWE, thereby impairing shear wave generation and propagation, which results in signal loss or unreliable measurement.
In our study, our proposed five color patterns demonstrated extremely high diagnostic ability (AUC =0.972) for the differentiation of BMUS-suspicious ALNs, offering a reliable preoperative assessment tool that could potentially reduce unnecessary diagnostic surgeries.
However, several limitations to this study should be acknowledged. First, BMUS was unable to detect all ALNs and evaluate complete ALNs status, and so the proposed five color patterns of SWE were only for applicable ALNs observed on BMUS. Second, SWE of ALNs was performed for each patient with suspected breast cancer in the study period, but a portion of patients with breast cancer did not undergo core needle biopsy and/or surgical biopsy; thus, selection bias might have been introduced. Third, although we employed a multicenter design, a larger sample size is needed to further validate our findings.

Conclusions
The proposed qualitative SWE patterns could provide reliable diagnostic discrimination of benign and metastatic BMUS-suspicious ALNs, which could optimize surgical planning, reduce overtreatment, and improve the quality of life of patients being treated for breast cancer.

Supplementary
The article’s supplementary files as