Abstract：
OBJECTIVE: This study aims to determine which qualitative and quantitative US features are independently associated with malignancy, including those derived from grayscale imaging morphology, shear wave elastography (SWE), and texture analysis.
METHODS: This single-center retrospective study was approved by the institutional research ethics board. Consecutive breast US studies performed between January and December 2020 were included. Images were acquired using a Canon Aplio i800 US unit (Canon Medical Systems, Inc., CA) and i18LX5 wideband linear matrix transducer. Grayscale US features, SWE mean, and median elasticity were obtained. Single representative grayscale images were analyzed using dedicated software (LIFEx, version 6.30). First-order and gray-level co-occurrence matrix second-order texture features were extracted. Multivariate logistic regression was performed to assess for predictors of malignancy (STATA v16.1).
RESULTS: One hundred forty-seven cases with complete SWE data were selected for analysis (mean age 54.3, range 21-92). The following variables were found to be independently associated with malignancy: age (P <.001), family history (P = .013), irregular mass shape (P = .024), and stiffness on SWE (mean SWE ≥40 kPa; P <.001). Remaining variables (including texture features) were not found to be independently associated with malignancy (P >.05).
CONCLUSIONS: US texture analysis features were not associated with malignancy independent of other qualitative and quantitative US characteristics currently utilized in clinical practice. This suggests texture analysis may not be warranted when differentiating benign and malignant breast masses on US. In contrast, irregular mass shape on grayscale imaging and increased stiffness on SWE were found to be independent predictors of malignancy.
METHODS: This single-center retrospective study was approved by the institutional research ethics board. Consecutive breast US studies performed between January and December 2020 were included. Images were acquired using a Canon Aplio i800 US unit (Canon Medical Systems, Inc., CA) and i18LX5 wideband linear matrix transducer. Grayscale US features, SWE mean, and median elasticity were obtained. Single representative grayscale images were analyzed using dedicated software (LIFEx, version 6.30). First-order and gray-level co-occurrence matrix second-order texture features were extracted. Multivariate logistic regression was performed to assess for predictors of malignancy (STATA v16.1).
RESULTS: One hundred forty-seven cases with complete SWE data were selected for analysis (mean age 54.3, range 21-92). The following variables were found to be independently associated with malignancy: age (P <.001), family history (P = .013), irregular mass shape (P = .024), and stiffness on SWE (mean SWE ≥40 kPa; P <.001). Remaining variables (including texture features) were not found to be independently associated with malignancy (P >.05).
CONCLUSIONS: US texture analysis features were not associated with malignancy independent of other qualitative and quantitative US characteristics currently utilized in clinical practice. This suggests texture analysis may not be warranted when differentiating benign and malignant breast masses on US. In contrast, irregular mass shape on grayscale imaging and increased stiffness on SWE were found to be independent predictors of malignancy.
摘要：
目的：本研究旨在确定哪些定性和定量US特征与恶性肿瘤独立相关，包括来自灰度成像形态学的那些，剪切波弹性成像（SWE），和纹理分析。
方法：这项单中心回顾性研究得到了机构研究伦理委员会的批准。纳入2020年1月至12月进行的连续乳腺美国研究。图像是使用佳能Aplioi800美国单位(佳能医疗系统，Inc.,CA)和i18LX5宽带线性矩阵换能器。美国的灰度特征，SWE的意思是，并获得了中值弹性。使用专用软件(LIFEx，版本6.30)。提取一阶和灰度共生矩阵二阶纹理特征。采用多因素logistic回归分析评估恶性肿瘤的预测因子（STATAv16.1）。
结果：选择了147例具有完整SWE数据的病例进行分析（平均年龄54.3，范围21-92）。发现以下变量与恶性肿瘤独立相关：年龄（P<0.001），家族史（P=.013），不规则的质量形状（P=.024），和SWE上的刚度（平均SWE≥40kPa；P<.001）。未发现其余变量（包括纹理特征）与恶性肿瘤独立相关（P>.05）。
结论：US纹理分析特征与恶性肿瘤无关，独立于目前临床实践中使用的其他定性和定量US特征。这表明在US上区分良性和恶性乳腺肿块时，可能不需要进行纹理分析。相比之下,灰阶成像上不规则的肿块形状和SWE上增加的硬度被发现是恶性肿瘤的独立预测因子.
方法：这项单中心回顾性研究得到了机构研究伦理委员会的批准。纳入2020年1月至12月进行的连续乳腺美国研究。图像是使用佳能Aplioi800美国单位(佳能医疗系统，Inc.,CA)和i18LX5宽带线性矩阵换能器。美国的灰度特征，SWE的意思是，并获得了中值弹性。使用专用软件(LIFEx，版本6.30)。提取一阶和灰度共生矩阵二阶纹理特征。采用多因素logistic回归分析评估恶性肿瘤的预测因子（STATAv16.1）。
结果：选择了147例具有完整SWE数据的病例进行分析（平均年龄54.3，范围21-92）。发现以下变量与恶性肿瘤独立相关：年龄（P<0.001），家族史（P=.013），不规则的质量形状（P=.024），和SWE上的刚度（平均SWE≥40kPa；P<.001）。未发现其余变量（包括纹理特征）与恶性肿瘤独立相关（P>.05）。
结论：US纹理分析特征与恶性肿瘤无关，独立于目前临床实践中使用的其他定性和定量US特征。这表明在US上区分良性和恶性乳腺肿块时，可能不需要进行纹理分析。相比之下,灰阶成像上不规则的肿块形状和SWE上增加的硬度被发现是恶性肿瘤的独立预测因子.
