Risk stratification system for sentinel lymph node metastasis in clinically node-negative early breast cancer with sonographically abnormal but cytologically negative axilla Authors: Wenjin Lin, Junrong Zhang, Rongxi Liang, Qin Ye, Zhenhu Lin, Xiujuan Zhang Published: 2025 ABSTRACT AbstractFor clinically node-negative (cN0) breast cancer patients with sonographically suspicious lymph nodes (iN+) but negative fine-needle aspiration cytology (N0(f)), sentinel lymph node biopsy (SLNB) remains the first-line approach for axillary staging. However, most of these patients ultimately demonstrate SLN-negative pathology, highlighting a critical need for a risk stratification to safely reduce unnecessary SLNB procedures. To address this unmet clinical challenge, we developed a preoperative risk stratification model integrating clinicopathological and ultrasonographic features. This prospective cohort study analyzed 369 consecutive iN+/N0(f) early breast cancer patients who underwent contrast-enhanced ultrasound (CEUS)-guided SLN fine-needle aspiration between September 2022 and December 2024. Multivariable logistic regression identified five independent predictors of SLN metastasis: cT2 staging, tumor grade III, conventional ultrasound type III, CEUS pattern III/IV, and abnormal sentinel lymphatic channel morphology. The resultant nomogram demonstrated good discrimination, with AUCs of 0.83 (95% CI: 0.77–0.89) and 0.82 (95% CI: 0.71–0.92) in training and temporal validation cohorts, respectively. The nomogram-based risk stratification system demonstrated good clinical utility: the very-low-risk subgroup exhibited only 2.6% SLN-positive incidence (2/76), achieving a negative predictive value of 97.4%, with no macrometastases observed in SLN-positive cases. This risk stratification system provides a clinically implementable method for personalized axillary management, improving preoperative identification of iN+/N0(f) patients likely to benefit from SLNB exemption.Keywords: Lymph node, Contrast-enhanced ultrasound, Sentinel lymph node biopsy, Breast cancerSubject terms: Cancer, Medical research, Oncology, Risk factors In newly diagnosed breast cancer, axillary lymph node (ALN) status remains a critical determinant of prognosis and therapeutic decision-making1. For clinically node-negative (cN0) patients, sentinel lymph node biopsy (SLNB) is the gold standard for ALN staging2. However, SLNB is not without risks: approximately 6% of patients develop lymphedema, 4% experience chronic pain, and 23% report long-term quality-of-life impairments3–6. Importantly, growing evidence suggests that SLNB serves primarily as a diagnostic procedure rather than a therapeutic intervention, prompting reevaluation of its necessity in selected cN0 populations6–10. This paradigm shift has catalyzed the pursuit of minimally invasive axillary assessment modalities that balance diagnostic accuracy with reduced procedural morbidity—a critical goal modern imaging studies. Among evolving modalities, contrast-enhanced ultrasound (CEUS)-guided sentinel lymph node (SLN) fine-needle aspiration (FNA) has emerged as a promising alternative. By dynamically mapping lymphatic drainage pathways and identifying SLNs through real-time microbubble enhancement, CEUS achieves superior nodal visualization compared to conventional ultrasound (US)11. This technique contributes to axillary staging in two key ways: first, by enabling precise localization of the SLN (with reported accuracy of 90–100%12,13), it minimizes the risk of sampling error during FNA; second, by characterizing nodal microvascular perfusion, it provides diagnostic information based on distinct enhancement patterns, which show high inter-observer agreement and can help stratify metastatic risk14. The imaging advantage of SLN visualization translates into improved biopsy accuracy, with CEUS-guided procedures achieving diagnostic accuracy of 80–90% and 100% specificity in recent studies15–17. Despite these advances, critical gaps persist: a meta-analysis revealed that CEUS-guided FNA exhibited a suboptimal negative predictive value (NPV) of approximately 85% and a sensitivity of 54%, indicating a considerable potential for false-negative errors12. These limitations underscore the unmet need for a risk stratification tool to identify patients in whom CEUS-guided FNA results reliably exclude nodal metastasis. To address this unmet need, we hypothesize that integrating clinicopathological variables with US-specific features can refine preoperative risk assessment. Our study pursues two sequential objectives: Firstly, to identify key determinants of SLN metastasis in patients with image abnormal (iN+) but CEUS-guided FNA negative (N0(f)) axilla; Secondly, to establish a nomograph-based risk stratification system to define a low-risk population (NPV >95.0%) for safely omitting SLNB. This strategy approach aligns with evolving precision oncology frameworks, prioritizing oncologic safety while preserving patient quality of life through axillary de-escalation. Materials and methods This study was carried out in line with the moral standards stipulated in the Helsinki Declaration. It was approved by the Institutional Review Board of Fujian Medical University Union Hospital, with written informed consent obtained from all participants. Study population This prospective cohort study enrolled consecutive patients with cN0 early invasive breast cancer undergoing CEUS-guided FNA at Fujian Medical University Union Hospital between September 2022 and December 2024. Inclusion criteria comprised: (1) clinical node-negative (cN0) status, defined as no palpable lymph nodes on physical examination of the axilla. (2) sonographically abnormal axilla (iN+), defined as the presence of lymph nodes meeting any of the following criteria: eccentric or concentric cortical thickening with a thickness > 2.5 mm; or complete loss of the fatty hilum18. (3) FNA cytologically negative (N0(f)), defined as the absence of malignant or atypical cells suspicious for malignancy across all obtained samples. Exclusion criteria included: (1) technical failure of FNA resulting in inadequate material; (2) breast cancer-related neoadjuvant therapy or radiotherapy already planned; (3) patients with bilateral or recurrent breast cancer; (4) pregnancy (Fig. 1). Of 369 enrolled participants, the cohort was divided into a training set (n = 281; September 2022-July 2024) and an independent temporal validation set (n = 88; August 2024-December 2024) based on chronological recruitment. Flowchart of included and excluded participants. US: ultrasound, FNA: fine-needle aspiration, SLN: sentinel lymph node. Clinicopathological characteristics Baseline clinicopathological characteristics were systematically captured for all N0(f) patients through case collection forms. Prospectively collected variables included: demographic characteristics (age at diagnosis, menopausal status, body mass index [BMI]), tumor morphological features (clinical T-stage [cT], tumor number, and anatomical location), as well as histopathological parameters including histological subtype, tumor grade, and molecular status (Ki-67 index, estrogen receptor [ER] status, progesterone receptor [PR] status, and HER2 status). Molecular subtyping was categorized into three Intrinsic subtypes: HR+/HER2-, HER2-enriched, and triple-negative subtypes. Axillary pathology outcomes were ascertained through surgical reports, detailing: sentinel/excised lymph node status (SLN + vs. SLN−), metastatic node count, and metastasis size (micrometastasis [≤ 2 mm] vs. macrometastasis [>2 mm]). Axillary nodal burden was stratified into four predefined prognostic categories: SLN−: no metastatic deposits; SLN+ (micro): micrometastasis; SLN+ (1–2): 1–2 macrometastases; SLN+ (≥ 3): ≥3 macrometastases. Ultrasound examination All breast and ALN evaluations were conducted using pre-set imaging protocols on two high-resolution ultrasound systems: the Supersonic Aixplorer® (Supersonic Imagine, France) and GE Logiq E11 (GE Healthcare, USA), both equipped with linear high-frequency transducers (4–15 MHz and 5–12 MHz, respectively). The US contrast agent (SonoVue®, Bracco Imaging) was reconstituted by mixing the lyophilized powder with 2 mL of sterile normal saline. Intradermal administration of 0.5 mL was performed at four periareolar injection sites (3, 6, 9, and 12 o’clock positions), with each injection followed by 10 s of massage to promote lymphatic dispersion. Image acquisition employed dual-mode contrast imaging technology, maintaining a low mechanical index (mechanical index: 0.06–0.12) to preserve microbubble integrity. Real-time visualization of contrast-enhanced SLC allowed sequential identification of the first lymph node within the axillary. The first enhanced lymph node was anatomically defined as the SLN and marked on the skin surface using a surgical marker. In cases of concurrent enhancement of multiple lymph nodes, priority was given to either the first enhanced node or the node exhibiting the thickest cortical dimension for subsequent biopsy. The following parameters were systematically recorded: SLN short-axis diameter, cortical thickness, conventional US types, CEUS enhancement patterns, and sentinel lymphatic channel (SLC) status. All US images were independently analyzed by two sonographers with > 10 years of breast imaging experience; discrepancies were resolved through consensus review by a senior breast radiologist (20 + years’ expertise). Suspicious target SLN were categorized into three conventional US types based on cortical thickness and hilum status: Type I: Cortical thickness ≤ 3.5 mm but > 2.5 mm with preserved fatty hilum; Type II: Cortical thickness ≤ 4.5 mm but > 3.5 mm with preserved hilum; Type III: Cortical thickness > 4.5 mm or complete loss of fatty hilum. Referring to previous study19 and our experiences, CEUS enhancement patterns were classified into four patterns: Pattern I: Homogeneous parenchymal enhancement; Pattern II: Characteristic heterogeneous enhancement subtypes (including cribriform, ring-like, half-moon, crescent, and truncated enhancement); Pattern III: Non-specific heterogeneous enhancement; Pattern IV: No enhancement. SLC status was further classified as either normal (continuous linear channels with regular branching) or abnormal (distorted/attenuated architecture, netted collateral vessels, or complete interruption), following the previous study11. CEUS-FNA and axillary surgery Under CEUS guidance, FNA was performed on the most suspicious sentinel lymph node using a 25-gauge needle (Leep®, China), ensuring four separate passes per target node. All cytology smears were immediately fixed in 95% ethanol for subsequent cytopathological analysis. Patients with N0(f) results proceeded to standard axillary staging via SLNB. Intraoperative frozen section analysis of SLNs was performed for immediate decision-making: completion ALND was executed if SLN metastasis (including micrometastasis and macrometastasis) was detected. Among the training cohort, 218 patients underwent SLNB alone, while 63 required subsequent ALND based on frozen section results. Notably, 2 cases with intraoperatively negative SLNB findings were reclassified as micrometastases upon final paraffin-embedded section review. Statistical analysis Statistical analyses were performed using SPSS version 25.0 (IBM Corp., Armonk, NY, USA) and R version 4.4.0 (2024-04-24), along with Storm Statistics Platform (www.medsta.cn/software). Categorical variables were compared using Pearson’s chi-square test or Fisher’s exact test, as appropriate. Continuous variables were presented as mean ± standard deviation or median with interquartile range after Shapiro–Wilk testing, and compared using the independent Student’s t-test or the Mann-Whitney U test, respectively. The endpoint of the analysis was the presence of SLN metastasis. To identify independent predictors, a two-step approach was employed. First, univariate analyses were conducted on all available clinicopathological and ultrasonographic variables. Variables associated with SLN metastasis at a significance level of p < 0.10 in the univariate analysis were considered candidates for the multivariable model. Subsequently, these candidate variables were entered into a multivariable binary logistic regression model using the Enter method to identify independent predictors and adjust for potential confounders. The results of the final multivariable model are presented as odds ratios (ORs) with 95% confidence intervals (CIs). The coefficients from the final multivariable logistic regression model were used to construct a visual nomogram for predicting the individual probability of SLN metastasis. The model’s performance was evaluated in both the training and temporal validation cohorts. Discrimination was quantified by the Area Under the Receiver Operating Characteristic Curve (AUC-ROC). Calibration was assessed using the Hosmer-Lemeshow goodness-of-fit test and visualized with a calibration plot, which plots the predicted probabilities against the observed frequencies of SLN metastasis. Clinical utility was evaluated using Decision Curve Analysis (DCA) to estimate the net benefit of the model across a range of threshold probabilities. Finally, based on the total points calculated from the nomogram, patients were categorized into four risk groups according to the quartile distribution of the risk scores in the training cohort. The SLN metastasis rate, NPV, and axillary tumor burden for each risk stratification were calculated. Results Clinicopathological features The prospective training cohort included 281 consecutive patients with N0(f) breast cancer. Final histopathological assessment identified SLN metastases in 65 cases, resulting in a false-negative rate (FNR) of 23.1% (65/281). Among these SLN + patients, metastatic burden was categorized as follows: SLN+ (micro) in 21.5% (14/65), SLN+ (1–2) in 43.1% (28/65), and high tumor burden (SLN+ (≥ 3)) in 35.4% (23/65). Comparative analysis between SLN + and SLN − groups revealed no statistically significant associations across multiple variables, including age, menopausal status, BMI, tumor location, multifocality, histological subtype, intrinsic molecular subtype, ER/PR status, and HER2 status (all p > 0.05). However, cT stage demonstrated significant predictive value, with higher cT stage (cT2) being overrepresented in the SLN + group (27.57% vs. 14.58%, p = 0.014). Borderline statistical trends were observed for tumor grade (p = 0.084) and Ki-67 index (p = 0.069), though these did not reach predefined significance thresholds (Table 1). Table 1Comparison of clinicopathological characteristics between the SLN + and SLN- groups in the training cohort.VariablesSLN+ (n = 65)SLN- (n = 216)StatisticP valueAge, years, M (IQR)48 (42, 55)50 (42, 57)Z=-1.420.156Menopausal status, n(%)0.786 Premenopausal51 (78.46)166 (76.85)χ2=0.07 Postmenopausal14(21.54)50 (23.15)BMI, kg/m2, n(%) <2557 (87.69)191 (88.43)χ2=0.030.872 ≥258 (12.31)25 (11.57)Tumor location, n(%) Areolar4 (6.15)15 (6.94)χ2=4.410.353 Upper outer24 (36.92)81 (37.50) Upper inner18 (27.69)68 (31.48) Lower outer16 (24.62)32 (14.81) Lower inner3 (4.62)20 (9.26)cT staging, n(%) T114 (21.54)82 (37.96)χ2=5.99 0.014 T251 (78.46)134 (62.04)Tumor number, n(%) Unifocal59 (90.77)205 (94.91)χ2=0.870.352 Multifocal6 (9.23)11 (5.09)Histologic subtype, n(%) Invasive ductal carcinoma62 (95.38)199 (92.13)–0.358 Invasive lobular carcinoma3 (4.62)9 (4.17) Others0 (0.00)8 (3.70)Intrinsic subtype, n(%) HER2+18 (27.69)58 (26.85)χ2=0.850.654 HR+/HER2−34 (52.31)103 (47.69) Triple-negative13 (20.00)55 (25.46)ER/PR status, n(%) Negative25 (38.46)101 (46.76)χ2=1.390.238 Positive40 (61.54)115 (53.24)HER2 status, n(%) Negative47 (72.31)158 (73.15)χ2=0.020.894 Positive18 (27.69)58 (26.85)Ki-67 index, n(%) ≤14%43 (66.15)167 (77.31)χ2=3.30 0.069 >14%22 (33.85)49 (22.69)Tumor grading, n(%) I9 (13.85)47 (21.76)χ2=4.95 0.084 II41 (63.08)141 (65.28) III15 (23.08)28 (12.96)BMI: body mass index, Z: Mann-Whitney test, χ2: Chi-square test, -: Fisher exact, M: Median, IQR: Interquartile Range.Significant values are in bold. Comparison of clinicopathological characteristics between the SLN + and SLN- groups in the training cohort. BMI: body mass index, Z: Mann-Whitney test, χ2: Chi-square test, -: Fisher exact, M: Median, IQR: Interquartile Range. Significant values are in bold. Ultrasound characteristics Univariate analysis of sonographic characteristics demonstrated significant associations between SLN metastasis status and three imaging features in the training cohort. Conventional US classification demonstrated good discriminative capacity (p < 0.001), with SLN + rates progressively increasing across categories: Type I showed 9.3% positivity (11/118), Type II 26.7% (35/131), and Type III 59.4% (19/32). CEUS patterns were similarly predictive (p < 0.001), displaying a hierarchical risk stratification: Pattern I had 6.0% SLN+ (5/83), Pattern II 13.7% (14/102), Pattern III 44.4% (32/72), and Pattern IV 58.3% (14/24). Abnormal SLC status also showed significant correlation with SLN metastasis (p = 0.009). However, nodal size showed no discriminative capacity between SLN + and SLN- groups (p > 0.05) (Table 2). Table 2Comparison of ultrasound characteristics between the SLN + and SLN- groups in the training cohort.VariablesSLN+ (n = 65)SLN- (n = 216)StatisticP valueNodal size, cm, Mean ± SD0.78 ± 0.480.91 ± 0.65t = 1.600.110US type, n(%) Type I11 (16.92)107 (49.54)χ2= 37.24 < 0.001 Type II35 (53.85)96 (44.44) Type III19 (29.23)13 (6.02)CEUS pattern, n(%) Pattern I5 (7.69)78 (36.11)χ2= 59.51 < 0.001 Pattern II14 (21.54)88 (40.74) Pattern III32 (49.23)40 (18.52) Pattern IV14 (21.54)10 (4.63)SLC, n(%)χ2= 6.83 0.009 Abnormal22 (33.85)40 (18.52) Normal43 (66.15)176 (81.48)t: t-test, χ2: Chi-square test, SD: standard deviation, US: ultrasound, CEUS: contrast-enhanced ultrasound, SLC: sentinel lymphatic channel.Significant values are in bold. Comparison of ultrasound characteristics between the SLN + and SLN- groups in the training cohort. t: t-test, χ2: Chi-square test, SD: standard deviation, US: ultrasound, CEUS: contrast-enhanced ultrasound, SLC: sentinel lymphatic channel. Significant values are in bold. These findings were validated in the independent cohort, where conventional US type (p < 0.001) and CEUS patterns (p = 0.001) remained robust predictors of SLN metastasis (Supplementary table S1). The consistency of these associations across cohorts underscores their clinical relevance for risk stratification in N0(f) patients. Nomogram prediction model construction Multivariable logistic regression analysis incorporating clinicopathological, and ultrasonographic variables identified five independent predictors of SLN metastasis in N0(f) patients (all p < 0.05, Table 3): cT2 staging (OR = 2.68), tumor grade III (OR = 3.68), conventional US type III (OR = 7.25), CEUS pattern (pattern III [OR = 9.77], pattern IV [OR = 4.99]), and abnormal SLC status (OR = 2.27). These predictors were integrated into a nomogram (Fig. 2) where individual risk contributions were quantified through weighted point assignments. Total risk scores, calculated by summing component scores, demonstrated strong correlation with metastatic probability. Table 3Multivariate logistic regression analysis of factors associated with SLN metastasis (n = 281).FactorsβP valueOR (95%CI)cT staging cT11.00 (Reference) cT20.98 0.014 2.68 (1.22 ~ 5.90)Tumor grading I1.00 (Reference) II0.370.4541.45 (0.55 ~ 3.86) III1.30 0.025 3.68 (1.18 ~ 11.51)Ki 67 index ≤14%1.00 (Reference) >14%0.700.0642.02 (0.96 ~ 4.26)US type Type I1.00 (Reference) Type II0.420.3821.52 (0.60 ~ 3.87) Type III1.98 0.002 7.25 (2.12 ~ 24.81)CEUS pattern Pattern I1.00 (Reference) Pattern II0.670.2441.96 (0.63 ~ 6.10) Pattern III2.28 < 0.001 9.77 (2.98 ~ 32.05) Pattern IV1.61 0.044 4.99 (1.04 ~ 23.88)SLC status Normal1.00 (Reference) Abnormal0.82 0.039 2.27 (1.04 ~ 4.94)OR: Odds Ratio, CI: Confidence Interval, US: ultrasound, CEUS: contrast-enhanced ultrasound, SLC: sentinel lymphatic channel.Significant values are in bold. Multivariate logistic regression analysis of factors associated with SLN metastasis (n = 281). OR: Odds Ratio, CI: Confidence Interval, US: ultrasound, CEUS: contrast-enhanced ultrasound, SLC: sentinel lymphatic channel. Significant values are in bold. Nomogram prediction model for SLN metastasis in N0(f) patients. Locate each predictor on its axis, drop a vertical line to the ‘Points’ axis, sum the points, and locate the total on ‘Total Points’ to read the predicted probability of SLN metastasis. CEUS: contrast-enhanced ultrasound, SLC: sentinel lymphatic channel. The model demonstrated good discriminative capacity, achieving AUC values of 0.83 (95%CI 0.77–0.89) in the training cohort and 0.82 (95%CI 0.71–0.92) in the validation cohort (Fig. 3). The calibration curves in Fig. 4 show good agreement between predicted and observed probabilities in both cohorts. Decision curve analysis (Fig. 5) revealed that the nomogram model consistently outperformed both “treat-all” and “treat-none” strategies across a wide range of threshold probabilities, demonstrating its superior clinical utility and significant value for practical application. Receiver operating characteristic curve of the nomogram for the training and validation cohorts, demonstrating good discriminative performance in both. Calibration curves of the nomogram in the training (A) and validation (B) cohorts. The bias-corrected curves assess predictive accuracy; closer alignment with the diagonal ideal line indicates superior performance. Good calibration was observed in both cohorts. Decision curve analysis of the nomogram. The model yields net clinical benefit across a broad spectrum of threshold probabilities, indicating its utility for guiding clinical interventions. Risk stratification system for SLN metastasis Using nomogram-derived risk scores, patients were stratified into four risk groups based on quartile thresholds: very-low-risk (<77 points, n = 76), low-risk (78–109 points, n = 71), medium-risk (110–176 points, n = 65), and high-risk (≥ 176 points, n = 69). SLN metastasis incidence demonstrated a strong risk-gradient relationship: 2.63% (2/76) in very-low-risk, 5.63% (4/71) in low-risk, 24.62% (16/65) in medium-risk, and 62.32% (43/69) in high-risk groups (Fig. 6A). Risk stratification system for N0(f) patients in the training cohort (A) and validation cohort (B). Tumor burden stratification revealed progressive escalation with risk category. High-risk patients exhibited 24.64% (17/69) of SLN+ (≥ 3 nodes), significantly exceeding rates in medium-risk (9.23%, 6/65), low-risk (0%), and very-low-risk (0%) groups. Importantly, macrometastases were absent in very-low-risk patients, while high tumor burden (≥ 3 nodes) was exclusively observed in medium- and high-risk cohorts. Using very-low-risk classification for SLNB exemption achieved 97.4% NPV (NPV: 97.4% [95% CI: 94.2–100%], sensitivity: 34% [28–41%], specificity: 97% [93–100%]), correctly identifying 34.3% (74/216) of true SLN- cases. Validation confirmed clinical utility, with no SLN + cases in very-low-risk cohort and 83.3% (5/6) of high tumor burden concentrated in high-risk group (Fig. 6B). SLNB omission in very-low-risk patients achieved perfect NPV (100% [95% CI: 100–100%]), safely excluding 29.0% (20/69) from surgical staging. Discussion Recent landmark trials, including INSEMA18 and SOUND20, have validated the safety of omitting axillary surgery in early-stage cT1N0 patients with HR+/HER2- tumors undergoing breast-conserving surgery and radiotherapy. These trials established a critical foundation for reducing surgical morbidity while maintaining disease control. However, their protocols relied on preoperative ultrasound-guided biopsy to exclude nodal involvement in iN + patients—a method with a FNR of 28.9% in similar populations21. At present, clinicians must carefully consider the impact of insufficient or absent SLN diagnosis on subsequent treatment decisions when selecting patients for axillary surgery exemption22. In this study, we developed a nomogram-based risk stratification system to stratify the risk of SLN metastasis in iN + but N0(f) patients. We found that in the very low-risk group, the risk of SLN + was low, with NPVs reaching over 95% and lower axillary tumor burden (only micrometastases) among SLN + cases. These findings will enhance clinicians’ confidence in identifying patients who may be eligible for exemption from SLNB. In our study, 23.13% of N0(f) patients were ultimately diagnosed with SLN positivity, and 8.19% (23/281) had high axillary tumor burden, which continues to present a challenge to breast surgeons’ decisions regarding axillary management. To more accurately identify SLN + cases, we used multivariate regression analysis to select five independent risk factors associated with SLN metastasis from clinical-pathological and ultrasound features: cT2 stage, tumor grade III, conventional US type III, CEUS pattern (patterns III and IV), and abnormal SLC status. Consistent with previous studies23–25, our analysis revealed a significant correlation between primary tumor size and the risk of SLN metastasis, further supporting tumor size as a reliable preoperative predictor for surgical planning in early breast cancer management. Besides, the study by Alsumai et al.25, which identified tumor grade as a significant predictor of SLNB positivity in cN0 breast cancer patients, aligns closely with our findings, underscoring the potential role of tumor grade in preoperative risk stratification for this patient population. Notably, about 50% of patients with US type III or CEUS patterns III/IV ultimately had SLN + disease, emphasizing the imperative for histological confirmation in these high-risk subgroups14,19. Conversely, the high NPV observed for Pattern I (94%) and Pattern II (86.3%) in our cohort aligns with existing literature19 and supports the role of CEUS in reliably identifying low-risk nodes. Abnormal SLC status—a sonographic manifestation of tumor-related lymphatic invasion—proved particularly significant, with a 2.27-fold higher SLN metastasis risk. This finding aligns with the biological premise that lymphatic architectural disruption facilitates nodal spread11. In the study, we stratified the axillary metastasis risk according to the quartile distribution of the nomogram-based scores among all N0(f) patients. In the very-low-risk group, the NPV for SLN metastasis reached 97.37%, and 34.26% of true SLN- cases could be identified, which is comparable to the diagnostic value reported in Morawitz J’s study26. A meta-analysis27 showed that the FNR for SLNB was 8.61% (95%CI: 8.05–9.2%), indicating that the risk stratification system for N0(f) patients has great potential to be an alternative to SLNB for very-low-risk patients with NPVs above 95.0%, providing a basis for identifying appropriate candidates for SLNB exemption. In recent years, relevant researches28,29 have focused on the use of ultrasound-guided needle biopsy to predict axillary tumor burden, helping with the decisions of neoadjuvant therapy and axillary surgery. In this study, the proportion of axillary tumor high-burden cases among N0(f) patients was low (8.19%), which is consistent with findings from previous study30. The very-low-risk group showed no cases of macrometastasis in either the training or validation cohorts, which can reduce the treatment burden for patients while minimizing the risk of undertreatment for SLNB exemption. In the low-risk group, more than 90% of patients exhibited SLN- results. Importantly, none of the SLN + cases showed a high axillary tumor burden, indicating that this subgroup may be eligible for ALND exemption, in accordance with the recently SENOMAC trial31. For the medium- and high-risk groups, there remained a high proportion of SLN + cases, especially in the high-risk group, making SLNB the standard method for ALN staging. Recently, radiomics and machine learning models have been developed to predict the axillary status in breast cancer patients non-invasively, with an accuracy of up to 97.0%32. This method could potentially be an effective tool for distinguishing SLN- from SLN + among patients in the medium- and high-risk groups. However, the generalizability of the artificial intelligence models in clinical applications still faces challenges. There are several limitations in this study. Firstly, the single-center design with limited sample size restricts generalizability, and the risk stratification model requires validation through large-scale, multi-institutional cohorts to confirm its diagnostic accuracy and clinical utility. Secondly, while most clinical-pathological, and ultrasound characteristics of CEUS-guided FNA were included in our study, its false-negative results may still be influenced by technical variability in operator proficiency, cytopathological processing, and interobserver interpretation—factors that are not systematically addressed in our current framework. Future research should consider incorporating these factors into the study. Conclusion This study demonstrates that cT2 staging, tumor grade III, US type III, CEUS patterns III/IV, and SLC abnormality are independent risk factors associated with SLN metastasis. A nomogram developed based on these five risk factors for predicting SLN metastasis showed good predictive performance. The nomogram-derived risk stratification system for SLN metastasis can assist in the axillary step-down management of breast cancer patients with N0(f) status. Especially in the very-low-risk group, the NPV for predicting SLN metastasis exceeded 95%. Moreover, the SLN + cases in the very-low-risk stratum had relatively low axillary tumor burden, enhancing clinician confidence in determining when SLNB can be safely omitted. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Author contributions W.J.L. and J.R.Z. convinced and designed the study, collected and analyzed data and contributed to the first draft of the manuscript. X.J.Z. and Z.H.L. revised the manuscript and reviewed data. R.X.L., Q.Y. participated in the data collection and management. All authors read and approved the final manuscript. Funding This work was supported by Joint Funds for the innovation of Science and Technology, Fujian province (2024Y9307). Data availability The datasets used and analyzed during the current study are available from the corresponding author on reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Wenjin Lin and Junrong Zhang contributed equally to this work. Contributor Information Zhenhu Lin, Email: 377579875@qq.com. Xiujuan Zhang, Email: 13075814040@163.com. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-025-28047-w. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.