目的 探讨乳腺肿块生长方位量化联合S-Detect技术对预测乳腺癌腋窝淋巴结（ALN）转移的价值。 方法 收集2023年3月至2024年10月于医院住院的163例乳腺癌患者资料，依据ALN病理结果分为转移组（n = 62）与未转移组（n = 101）。所有患者术前行常规超声及S-Detect检查。采用单因素和多因素回归分析各观察指标与ALN转移的相关性，筛选出有意义的指标并建立logistic回归预测模型，采用受试者工作特征（ROC）曲线评价该模型的预测价值。 结果 单因素分析显示，肿块的最大径、边界、边缘、钙化、方位角、血流在两组间的差异有统计学意义（P < 0.05）。多因素分析结果显示钙化、边界、方位角、边缘、最大径是预测ALN状态的独立危险因素（P < 0.05）。依此构建的logistic回归预测模型：Y = -7.995 + 2.299 × 最大径 + 1.171 × 边界 + 2.137 × 边缘 + 1.397 × 钙化 + 0.034 × 方位角。该联合预测模型的AUC为0.869，均大于各独立影响因素的AUC（P < 0.05），联合预测模型与病理结果的一致性良好（Kappa = 0.701，P < 0.05）。 结论 量化乳腺肿块的方位角有助于预测ALN转移，并增强对非平行取向的解释和应用。乳腺肿块生长方位量化联合S-Detect技术对乳腺癌ALN转移具有较好的预测价值，可以给个性化治疗提供参考依据。

Objective To investigate the utility of combining breast mass growth orientation quantification with the S-Detect technique for predicting axillary lymph node （ALN） metastasis in breast cancer. Methods Data was collected from 163 breast cancer patients admitted to our hospital between March 2023 and October 2024， who were categorized into metastatic （n = 62） and non-metastatic （n = 101） groups based on ALN pathology results. All patients underwent routine preoperative ultrasound and S-Detect examination. Univariate and multivariate regression analyses were performed to assess the correlation between each observational index and ALN metastasis. Significant indexes were identified through screening， leading to the establishment of a logistic regression prediction model. The predictive value of the model was evaluated using receiver operating characteristic （ROC） curve analysis. Results The univariate analysis revealed statistically significant differences （P < 0.05） in the maximum diameter of the mass， border characteristics， margin features， calcification patterns， orientation angle， and blood flow between the two groups. Multifactorial analysis demonstrated that calcification， border characteristics， orientation angle， margin features， and maximum diameter independently influenced the prediction of axillary lymph node （ALN） status in breast cancer patients （P < 0.05）. Consequently， a logistic regression prediction model was constructed as follows： Y = -7.995 + 2.299 × maximal diameter + 1.171 × border + 2.137 × margin + 1.397 × calcication + 0.034 × orientation angle. The area under curve （AUC） for this combined prediction model was 0.869 which significantly outperformed each independent influencing factor alone （P < 0.05）， indicating good agreement between this joint prediction model and pathological results （Kappa = 0.701， P < 0.05）. Conclusions Quantification of the orientation angle of a breast mass aids in predicting axillary lymph node （ALN） metastasis and enhances the interpretation and application of non-parallel orientations. The combination of quantifying growth orientation based on breast mass with artificial intelligence S-Detect technique demonstrates promising predictive value for ALN metastasis in breast cancer， providing a reference basis for personalized treatment.