The purpose of this study was to determine whether incorporation of the ultrasound (US) features of the primary tumor and axillary lymph node (ALN) could improve the prediction of high axillary nodal burden (HNB) and, thus, avoid unnecessary sentinel lymph node biopsy (SLNB). A total of 347 patients with Breast Imaging Reporting and Data System US category 4 or 5 breast cancer lesions were included. Their pre-operative US features and post-operative pathologic results were collected. The patients were then divided into the following groups based on surgical histology: limited nodal burden (LNB: 0–2 LNs involved) and heavy nodal burden (HNB: ≥3 metastatic LNs). Univariate and multivariate logistic regression analyses were conducted to determine the most valuable variables for HNB prediction. Receiver operating characteristic curves were obtained to assess their values. We found that a non-circumscribed margin, cortical thickness (≥3 mm) and number (≥3) of suspicious ALNs are indicators for HNB prediction. The false-negative rate (FNR) in model 1 (cortical thickness + number of suspicious ALNs) was 15.5% versus 3.4% in model 2 (non-circumscribed margin + cortical thickness + number of suspicious ALNs). Our results indicate that combining the US features of the primary tumor and ALNs can reduce the FNR during HNB prediction.

本研究的目的是确定合并原发肿瘤和腋窝淋巴结 (ALN) 的超声 (US) 特征是否可以提高高腋窝淋巴结负荷 (HNB) 的预测，从而避免不必要的前哨淋巴结活检(SLNB)。共有 347 名乳房影像报告和数据系统美国 4 类或 5 类乳腺癌病变患者被纳入研究。收集他们的术前超声特征和术后病理结果。然后根据手术组织学将患者分为以下组：有限的淋巴结负荷（LNB：涉及 0-2 个 LN）和重型淋巴结负荷（HNB：≥3 个转移性淋巴结）。进行单变量和多变量逻辑回归分析以确定对 HNB 预测最有价值的变量。获得接受者操作特征曲线以评估它们的值。我们发现非限制性边缘、皮质厚度 (≥3 mm) 和可疑 ALN 的数量 (≥3) 是 HNB 预测的指标。模型 1（皮质厚度 + 可疑 ALN 数量）中的假阴性率 (FNR) 为 15.5%，而模型 2（非限制边缘 + 皮质厚度 + 可疑 ALN 数量）为 3.4%。我们的结果表明，结合原发肿瘤和 ALN 的 US 特征可以降低 HNB 预测过程中的 FNR。模型 2 中的 4%（非限制性边缘 + 皮质厚度 + 可疑 ALN 的数量）。我们的结果表明，结合原发肿瘤和 ALN 的 US 特征可以降低 HNB 预测过程中的 FNR。模型 2 中的 4%（非限制性边缘 + 皮质厚度 + 可疑 ALN 的数量）。我们的结果表明，结合原发肿瘤和 ALN 的 US 特征可以降低 HNB 预测过程中的 FNR。