Ultrasound‐Derived Nomogram for Differentiating <scp>HER2</scp> +/ <scp>HER2</scp> − Subtypes in Luminal B‐Type Breast cancer: Principal Ultrasound Determinants and Model Validation

ABSTRACT Objective To identify ultrasound predictors of HER2 status in Luminal B‐type breast cancer and develop an interpretable preoperative nomogram. Methods This retrospective study included 159 patients divided into training ( n = 111) and validation ( n = 48) cohorts. Candidate predictors were screened by univariable logistic regression ( p &lt; 0.20), followed by multivariable logistic regression with backward AIC‐based selection. LASSO regression, bootstrap internal validation, ROC analysis, calibration, and DCA were performed. Results Luminal B (HER2+) and Luminal B (HER2−) tumors differed in orientation, spiculation, hyperechoic halo, microcalcification, Adler grade blood flow, and posterior echo patterns. The final model retained orientation, spiculation, hyperechoic halo, Adler grade blood flow, and posterior echo attenuation; LASSO selected the same predictors. The nomogram showed good discrimination (training AUC, 0.827; validation AUC, 0.800). Bootstrap validation yielded optimism‐corrected AUC and calibration slope values of 0.801 and 0.728, respectively. Calibration showed acceptable agreement, and DCA suggested potential clinical utility. Conclusion The ultrasound‐based nomogram may provide a noninvasive, interpretable aid to preoperative HER2 risk stratification in Luminal B‐type breast cancer, but prospective multicenter validation is required before broader clinical application.