PDF(Pubmed)
Abstract：
UNASSIGNED: This study aimed to investigate whether quantitative radiomics features extracted from conventional ultrasound (CUS) and contrast-enhanced ultrasound (CEUS) of primary breast lesions can help noninvasively predict axillary lymph nodes metastasis (ALNM) in breast cancer patients.
UNASSIGNED: A total of 111 breast cancer patients with 111 breast lesions were prospectively enrolled. All the included patients received presurgical CUS screening and CEUS examination and were randomly assigned to the training and validation sets at a ratio of 7:3 (n = 78 versus 33). Radiomics features were respectively extracted based on CUS and CEUS using the PyRadiomics package. The max-relevance and min-redundancy (MRMR) and least absolute shrinkage and selection operator (LASSO) analyses were used for feature selection and radiomics score calculation in the training set. The variance inflation factor (VIF) was performed to check the multicollinearity among selected predictors. The best performing model was selected to develop a nomogram using binary logistic regression analysis. The calibration and clinical utility of the nomogram were assessed.
UNASSIGNED: The model combining CUS reported ALN status, CUS radiomics score (CUS-radscore) and CEUS radiomics score (CEUS-radscore) exhibited the best performance. The areas under the curves (AUC) of our proposed nomogram in the training and external validation sets were 0.845 [95% confidence interval (CI), 0.739-0.950] and 0.901 (95% CI, 0.758-1). The calibration curves and decision curve analysis (DCA) demonstrated the nomogram\'s robust consistency and clinical utility.
UNASSIGNED: The established nomogram is a promising prediction tool for noninvasive prediction of ALN status. The radiomics features based on CUS and CEUS can help improve the predictive performance.
UNASSIGNED: A total of 111 breast cancer patients with 111 breast lesions were prospectively enrolled. All the included patients received presurgical CUS screening and CEUS examination and were randomly assigned to the training and validation sets at a ratio of 7:3 (n = 78 versus 33). Radiomics features were respectively extracted based on CUS and CEUS using the PyRadiomics package. The max-relevance and min-redundancy (MRMR) and least absolute shrinkage and selection operator (LASSO) analyses were used for feature selection and radiomics score calculation in the training set. The variance inflation factor (VIF) was performed to check the multicollinearity among selected predictors. The best performing model was selected to develop a nomogram using binary logistic regression analysis. The calibration and clinical utility of the nomogram were assessed.
UNASSIGNED: The model combining CUS reported ALN status, CUS radiomics score (CUS-radscore) and CEUS radiomics score (CEUS-radscore) exhibited the best performance. The areas under the curves (AUC) of our proposed nomogram in the training and external validation sets were 0.845 [95% confidence interval (CI), 0.739-0.950] and 0.901 (95% CI, 0.758-1). The calibration curves and decision curve analysis (DCA) demonstrated the nomogram\'s robust consistency and clinical utility.
UNASSIGNED: The established nomogram is a promising prediction tool for noninvasive prediction of ALN status. The radiomics features based on CUS and CEUS can help improve the predictive performance.
摘要：
■本研究旨在探讨从常规超声（CUS）和超声造影（CEUS）提取的乳腺原发灶的定量影像组学特征是否有助于无创性预测乳腺癌患者的腋窝淋巴结转移（ALNM）。
■前瞻性纳入111例乳腺癌患者和111例乳腺病变。所有纳入的患者都接受了术前CUS筛查和CEUS检查，并以7:3的比例随机分配到训练和验证组(n=78对33)。使用PyRadiomics软件包分别基于CUS和CEUS提取Radiomics特征。最大相关性和最小冗余（MRMR）和最小绝对收缩和选择算子（LASSO）分析用于训练集中的特征选择和影像组学得分计算。执行方差膨胀因子(VIF)以检查所选预测因子之间的多重共线性。选择性能最佳的模型以使用二元逻辑回归分析来开发列线图。评估列线图的校准和临床实用性。
■组合CUS的模型报告了ALN状态，CUS影像组学评分（CUS-radscore）和CEUS影像组学评分（CEUS-radscore）表现最佳。训练和外部验证集中我们提出的列线图的曲线下面积（AUC）为0.845[95％置信区间（CI），0.739-0.950]和0.901（95%CI，0.758-1）。校准曲线和决策曲线分析（DCA）证明了列线图的稳定性和临床实用性。
■建立的列线图是用于ALN状态的非侵入性预测的有前途的预测工具。基于CUS和CEUS的影像组学功能有助于提高预测性能。
■前瞻性纳入111例乳腺癌患者和111例乳腺病变。所有纳入的患者都接受了术前CUS筛查和CEUS检查，并以7:3的比例随机分配到训练和验证组(n=78对33)。使用PyRadiomics软件包分别基于CUS和CEUS提取Radiomics特征。最大相关性和最小冗余（MRMR）和最小绝对收缩和选择算子（LASSO）分析用于训练集中的特征选择和影像组学得分计算。执行方差膨胀因子(VIF)以检查所选预测因子之间的多重共线性。选择性能最佳的模型以使用二元逻辑回归分析来开发列线图。评估列线图的校准和临床实用性。
■组合CUS的模型报告了ALN状态，CUS影像组学评分（CUS-radscore）和CEUS影像组学评分（CEUS-radscore）表现最佳。训练和外部验证集中我们提出的列线图的曲线下面积（AUC）为0.845[95％置信区间（CI），0.739-0.950]和0.901（95%CI，0.758-1）。校准曲线和决策曲线分析（DCA）证明了列线图的稳定性和临床实用性。
■建立的列线图是用于ALN状态的非侵入性预测的有前途的预测工具。基于CUS和CEUS的影像组学功能有助于提高预测性能。
