18F-FDG PET/CT影像组学判断乳腺癌人表皮生长因子受体2表达状态

目的 观察¹⁸F-FDG PET/CT影像组学判断乳腺癌人表皮生长因子受体2（HER2）表达状态的价值。方法 纳入100例乳腺癌患者，包括HER2（+）组28例、HER2（-）组72例，比较组间PET/CT参数，包括病灶最大标准摄取值（SUV_max）及其标准差（SD）、平均标准摄取值（SUV_mean）及肿瘤代谢体积的差异。于标准化后的PET图像中手动分割病灶ROI，获取其影像组学特征，根据降维后的影像组学特征及其系数的线性加权获得病灶影像组学风险评分（RRS）；针对差异有统计学意义的指标绘制受试者工作特征曲线，计算曲线下面积（AUC），评价其判断乳腺癌患者HER2状态的效能。采用Bootstrap 1000重复抽样进行内部验证，计算校正AUC，并以DeLong检验比较。以决策曲线分析评价患者净获益情况。结果 HER2（+）组病灶SUV_max、SUV_mean及SD均大于HER2（-）组（P均<0.05）。共获取704个影像组学特征，经筛选最终获得10个非零系数的特征用于计算RRS，HER2（+）组RRS大于HER2（-）组（P<0.001）。以RRS判断乳腺癌患者HER2状态的AUC大于病灶SUV_max、SUV_mean及SD （P均<0.05），其诊断敏感度为78.57%、特异度为77.78%。决策曲线分析结果表明，RRS可在较大概率阈值范围内带给患者更多净获益。结论 ¹⁸F-FDG PET/CT影像组学用于判断乳腺癌HER2状态具有一定价值。

18F-FDG PET/CT radiomics for judging human epidermal growth factor receptor-2 expression status of breast cancer

Objective To explore the value of ¹⁸F-FDG PET/CT radiomics for judging the expression status of human epidermal growth factor receptor-2 (HER2) of breast cancer. Methods Totally 100 patients with breast cancer were enrolled, including 28 cases in HER2 (+) group and 72 cases in HER2 (-) group. PET/CT parameters, including the maximum standard uptake value (SUV_max) and its standard deviation (SD), the mean standard uptake value (SUV_mean) and the metabolic tumor volumes were compared between groups. ROI of lesions were manually segmented on standardized PET images, and the radiomics features were acquired. The radiomic risk score (RRS) was obtained according to the linear weighting of the dimensional-reduced radiomics features and coefficients. Receiver operating characteristic curves of indexes being statistically different between groups were drawn, and area under the curves (AUC) was calculated to evaluate the efficacy of indicators for judging HER2 status of breast cancer. Bootstrap 1000 repeated sampling was used for internal validation, and the corrected AUC was calculated, and the corrected AUC was compared with DeLong test. The net patient benefit was evaluated using decision curve analysis (DCA). Results SUV_max, SUV_mean and SD in HER2 (+) group were significantly higher than those in HER2 (-) group (all P<0.05). A total of 704 radiomics features were obtained. After screening, 10 non-zero coefficient features were finally selected for calculating RRS. RRS in HER2 (+) group was greater than that in HER2 (-) group (P<0.001). AUC of RRS for judging HER2 status of breast cancer was greater than that of SUV_max, SUV_mean and SD (all P<0.05), with diagnostic sensitivity of 78.57% and specificity of 77.78%. DCA results showed that RRS could make higher net patient benefit within a larger probability threshold range. Conclusion ¹⁸F-FDG PET/CT radiomics had certain value for judging HER2 status of breast cancer.