From the Institute of Biomedical Engineering and Instrumentation, Hangzhou Dianzi University, High Education Zone, Hangzhou 310018, Zhejiang, China (M.F., Y.C., L. Li); Department of Radiology, Fudan University Shanghai Cancer Center, Shanghai, China (C.Y., L. Liu, Y.G., W.P., Q.B.); and Computational Bioscience Research Center, Computer, Electrical and Mathematical Sciences and Engineering Division, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia (X.G.).
Radiology. 2022 Mar;302(3):516-524. doi: 10.1148/radiol.2021210738. Epub 2021 Nov 30.
Background Radiogenomics explores the association between imaging features and genomic assays to uncover relevant prognostic features; however, the prognostic implications of the derived signatures remain unclear. Purpose To identify preoperative radiogenomic signatures of estrogen receptor-positive breast cancer associated with the Oncotype DX recurrence score (RS) and to evaluate whether they are biomarkers for survival and responses to neoadjuvant chemotherapy (NACT). Materials and Methods In this retrospective multicohort study, three data sets were analyzed. The radiogenomic development data set, with preoperative dynamic contrast-enhanced MRI and RS data obtained between January 2016 and October 2019 was used to identify radiogenomic signatures. Prognostic implications of the imaging signatures were assessed by measuring overall survival and recurrence-free survival in the prognostic assessment data set using a multivariable Cox proportional hazards model. The therapeutic implication of the radiogenomic signatures was evaluated by determining their ability to predict the response to NACT using the treatment assessment data set obtained between August 2015 and March 2019. Prediction performance was estimated by using the area under the receiver operating characteristic curve (AUC). Results The final cohorts included a radiogenomic development data set with 130 women (mean age, 52 years ± 10 [standard deviation]), a prognostic assessment data set with 116 women (mean age, 48 years ± 9), and a treatment assessment data set with 135 women (mean age, 50 years ± 11). Radiogenomic signatures ( = 11) of texture and morphologic and statistical features were identified to generate the predicted RS ( = 0.33, < .001). A predicted RS greater than 29.9 was associated with poor overall and recurrence-free survival ( = .001 and = .007, respectively); predicted RS was greater in women with a good NACT response (30.51 ± 6.92 vs 27.35 ± 4.04 [responders vs nonresponders], = .001). By combining the predicted RS and complementary features, the model achieved improved performance in prediction of the NACT response (AUC, 0.85; < .001). Conclusion Radiogenomic signatures associated with genomic assays provide markers of prognosis and treatment in estrogen receptor-positive breast cancer. © RSNA, 2021
背景 放射基因组学探讨了影像学特征与基因组检测之间的关联，以揭示相关的预后特征；然而，衍生特征的预后意义尚不清楚。目的 确定与 Oncotype DX 复发评分（RS）相关的雌激素受体阳性乳腺癌的术前放射基因组特征，并评估其是否为生存和对新辅助化疗（NACT）反应的生物标志物。材料与方法 在这项回顾性多队列研究中，分析了三个数据集。使用 2016 年 1 月至 2019 年 10 月期间获得的术前动态对比增强 MRI 和 RS 数据的放射基因组开发数据集来识别放射基因组特征。通过多变量 Cox 比例风险模型在预后评估数据集中测量总生存期和无复发生存期来评估成像特征的预后意义。通过使用 2015 年 8 月至 2019 年 3 月获得的治疗评估数据集来确定放射基因组特征预测 NACT 反应的能力，评估放射基因组特征的治疗意义。使用接收者操作特征曲线下面积（AUC）来估计预测性能。结果 最终的队列包括放射基因组开发数据集（130 名女性，平均年龄 52 岁±10[标准差]）、预后评估数据集（116 名女性，平均年龄 48 岁±9）和治疗评估数据集（135 名女性，平均年龄 50 岁±11）。确定了纹理、形态和统计特征的放射基因组特征（ = 11），以生成预测的 RS（ = 0.33，<.001）。预测的 RS 大于 29.9 与整体和无复发生存不良相关（ =.001 和 =.007，分别）；预测 RS 在 NACT 反应良好的女性中更高（30.51 ± 6.92 与 27.35 ± 4.04[应答者与非应答者]， =.001）。通过结合预测的 RS 和补充特征，该模型在预测 NACT 反应方面实现了更好的性能（AUC，0.85；<.001）。结论 与基因组检测相关的放射基因组特征为雌激素受体阳性乳腺癌提供了预后和治疗标志物。
