基于DCE-MRI影像组学特征联合ADC值预测乳腺癌Ki-67表达水平

Imaging radiomics features based on DCE-MRI combined with ADC in predicting expression level of Ki-67 in breast cancer

目的探讨动态对比增强MRI(dynamic contrast-enhanced MRI,DCE-MRI)影像组学特征联合表观扩散系数(apparent diffusion coefficient,ADC)值预测乳腺癌Ki-67表达水平的临床价值。材料与方法回顾性分析2018年12月至2021年12月间经病理证实的234例乳腺癌患者MRI影像资料,依据术后免疫组化结果,将其分为Ki-67高表达组(n=180)和低表达组(n=54),采用半自动分割的方式从DCE-MRI第1期增强图像中提取瘤体1906个组学特征,采用组内相关系数(intra-class correlation coefficient,ICC)、特征间线性相关性分析和最小绝对收缩与选择算子(least absolute shrinkage and selection operator,LASSO)最终筛选出4个最优特征构建影像组学模型,采用受试者工作特征(receiver operating characteristic,ROC)曲线评估影像组学、平均ADC值及二者联合模型的诊断效能。并使用校准曲线及决策曲线评价预测模型的临床实用性。结果从瘤体提取1906个特征,ICC分析剔除207个、特征间线性相关性分析剔除1626个,剩余73个特征LASSO降维筛选出4个最优组学特征。最终筛选出的4个组学特征,平均ADC值在两组间差异均有统计学意义(P<0.05)。影像组学、平均ADC值及联合模型预测Ki-67高表达的曲线下面积(area under the curve,AUC)分别为0.820、0.676和0.856,三者间的差异均有统计学意义(P<0.05),联合模型对Ki-67高表达的预测效能最佳,其AUC、敏感度和特异度分别为0.856、88.3%和74.1%,校准曲线及决策曲线显示联合模型具有临床应用价值。结论基于DCE-MRI组学特征联合平均ADC值对乳腺癌Ki-67高表达具有较高的预测效能,联合模型优于影像组学模型及平均ADC值。

Objective:To investigate the clinical value of imaging radiomics features based on dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI)combined with apparent diffusion coefficient(ADC)in predicting the expression level of Ki-67 in breast cancer.Materials and Methods:MRI images of 234 patients with breast cancer confirmed by pathology from December 2018 to December 2021 were retrospectively analyzed.According to postoperative immunohistochemical results,the tumors were divided into the Ki-67 high expression group(n=180)and low expression group(n=54).1906 radiomics features were extracted form the first phase of the DCE-MRI by semi-automatic separation method.Using intraclass correlation coefficient(ICC),the linear correlation analysis and the least absolute shrinkage and selection operator(LASSO),four features were selected to construct the radiomics model.Receiver operating characteristic(ROC)curves were used to evaluate the diagnostic effectiveness of the radiomics,average ADC values and combined models.Calibration curves and decision curves were used to evaluate the clinical usefulness of the predictive model.Results:A total of 1906 features were extracted from the tumor body,207 features were excluded by ICC analysis,1626 features were excluded by linear correlation analysis,and the remaining 73 features were selected by LASSO dimensionality reduction to select 4 optimal omics features.Four radiomics features and the average ADC values were significantly different between two groups(P<0.05).Radiomics model,the average ADC value and the combined model predicted that the area under the curve(AUC)of Ki-67 high expression were 0.820,0.676 and 0.856,respectively,with statistically significant differences each other(P<0.05).The combined model had the best predictive efficiency for Ki-67 expression,and its AUC,sensitivity and specificity were 0.856,88.3%and 74.1%,calibration curves and decision curves showed that the combined model had clinical application value.Conclusions:The combined model which constructed by the images radiomics features based on DCE-MRI and the average ADC values has high efficacy in predicting Ki-67 expression in breast cancer.The combined model is superior to the radiomics model and the average ADC value.