【摘要】 目的评价MRI纹理分析预测乳腺癌腋窝淋巴结(ALN)转移的价值。方法以172例乳腺癌为训练组,分为ALN阳性亚组(n=79)和阴性亚组(n=93),提取增强MRI及ADC图中ALN纹理特征,比较2亚组间纹理特征差异;筛选纹理特征,构建多因素Logistic回归模型,并对模型进行内部验证。以另外37例乳腺癌为测试组,行模型外部验证。结果基于增强MRI和ADC图选取16个纹理参数,其中熵的预测效能最优,曲线下面积(AUC)分别为0.781和0.786。经Lasso回归筛选11个纹理特征,以多因素Logistic回归建立预测模型,内部验证结果显示其AUC为0.906,用于测试组时AUC为0.859,预测效能良好。结论 MRI纹理分析预测乳腺癌ALN转移具有较高效能。更多还原

【Abstract】 Objective To explore the value of MRI texture analysis for predicting axillary lymph node(ANL) metastasis of breast cancer. Methods A total of 172 patients with pathologically proved breast cancer were enrolled as training group. According to the condition of ALN, the patients were divided into positive subgroup(n=79) and negative subgroup(n=93). The texture features of ALN on enhanced MRI and ADC images were extracted and compared between subgroups. The texture features were screened, and the multi-factor Logistic regression model was constructed, and the model was validated internally. Then other 37 breast cancer patients were taken as test group for external validation of the above model. Results Sixteen texture parameters were obtained based on enhanced MRI and ADC images. The prediction efficiency of entropy was the best, and the area under the curve(AUC) was 0.781 and 0.786, respectively. Eleven texture features were screened with Lasso regression, and the multi-factor Logistic regression model showed good prediction efficiency by internal and external verification, with AUC of 0.906 and 0.859, respectively, both indicating good prediction efficiency. Conclusion MRI texture analysis was high effective in predicting breast cancer ALN metastasis.更多还原