超声影像组学模型对乳腺癌人表皮生长因子受体2表达的预测价值

Construction of an ultrasound radiomics model for prediction of human epidermal growth factor receptor 2 expression in breast cancer

目的 探讨超声影像组学模型对乳腺癌分子标志物人表皮生长因子受体2(HER-2)表达的预测价值。方法 回顾2018年1月至2023年4月三峡大学第一临床医学院经病理检查证实为乳腺癌女性患者235例,其中HER-2表达阳性患者85例,HER-2表达阴性患者150例。将患者按7∶3随机分为训练集164例和验证集71例。选取显示乳腺病灶最大径线切面的超声图像,利用3D-Slicer软件手动图像分割和影像组学特征提取。然后对特征进行筛选,将特征进行Z-Score标准化,通过Pearson相关系数最小冗余最大相关算法和最小绝对收缩和选择算子筛选特征,利用AdaBoost算法构建影像组学模型。使用ROC曲线评估模型效能,Delong检验评价训练集和验证集间AUC的差异。结果 最终选取8个影像组学特征用于构建超声影像组学模型。该模型在训练集和验证集中预测HER-2阳性表达的AUC、准确度、灵敏度、特异度、阳性预测值和阴性预测值分别为0.830(95%CI:0.772~0.888)、0.718、0.853、0.642、0.574、0.885,0.771(95%CI:0.629~0.912)、0.723、0.706、0.733、0.600、0.815。Delong检验结果显示训练集和验证集AUC的差异无统计学意义(P>0.05)。结论 超声影像组学模型对乳腺癌分子标志物HER-2表达的预测具有重要价值。

Objective To explore the value of ultrasound radiomics model in predicting human epidermal growth factor receptor 2(HER-2)expression in breast cancer.Methods A total of 235 female patients with pathologically confirmed breast cancer at the First College of Clinical Medical Science of China Three Gorges University from January 2018 to April 2023 were retrospectively analyzed,including 85 cases with positive HER-2 expression and 150 cases with negative HER-2 expression.The patients were randomly divided into a training set(n=164)and a validation set(n=71)in a 7∶3 ratio.Ultrasound images depicting the largest diameter section of breast lesions were selected for analysis.The 3D-Slicer software was utilized to manually segment the images and extract radiomics features.Subsequently,the features were Z-Score normalized and filtered using the Pearson correlation coefficient,minimum redundancy maximum correlation algorithm,and minimum absolute shrinkage selection operator.Finally,the AdaBoost algorithm was employed to construct a radiomics model.The model performance was evaluated using the ROC curve,and the difference in the area under the ROC curve(AUC)between the training set and the validation set was evaluated using the Delong test.Results Eight radiomics features were selected to create ultrasound radiomics model.In the training set and validation set,the AUC,accuracy,sensitivity,specificity,positive predictive value,and negative predictive value of this model in predicting HER-2 positive expression were 0.830(95%CI:0.772-0.888)、0.718、0.853、0.642、0.574、0.885,0.771(95%CI:0.629-0.912)、0.723、0.706、0.733、0.600、0.815,respectively.The Delong test results indicated that there was no statistically significant difference in AUC between the training set and the validation set(P>0.05).Conclusion Ultrasound radiomics signatures may be used for predicting the expression status of breast cancer molecular marker HER-2.