Feng Zhan, Zhang Lixia, Qi Zhong, Shen Qijun, Hu Zhengyu, Chen Feng
Department of Radiology, College of Medicine, The First Affiliated Hospital, Zhejiang University, Hangzhou, China.
Department of Radiology, Hangzhou First People's Hospital, Hangzhou, China.
Front Oncol. 2020 Feb 28;10:279. doi: 10.3389/fonc.2020.00279. eCollection 2020.
To evaluate the potential application of computed tomography (CT) radiomics in the prediction of BRCA1-associated protein 1 () mutation status in patients with clear-cell renal cell carcinoma (ccRCC). In this retrospective study, clinical and CT imaging data of 54 patients were retrieved from The Cancer Genome Atlas-Kidney Renal Clear Cell Carcinoma database. Among these, 45 patients had wild-type and nine patients had mutation. The texture features of tumor images were extracted using the Matlab-based IBEX package. To produce class-balanced data and improve the stability of prediction, we performed data augmentation for the mutation group during cross validation. A model to predict mutation status was constructed using Random Forest Classification algorithms, and was evaluated using leave-one-out-cross-validation. Random Forest model of predict mutation status had an accuracy of 0.83, sensitivity of 0.72, specificity of 0.87, precision of 0.65, AUC of 0.77, F-score of 0.68. CT radiomics is a potential and feasible method for predicting mutation status in patients with ccRCC.
评估计算机断层扫描（CT）影像组学在预测透明细胞肾细胞癌（ccRCC）患者中乳腺癌1号关联蛋白1（）突变状态的潜在应用。在这项回顾性研究中，从癌症基因组图谱 - 肾透明细胞癌数据库中检索了54例患者的临床和CT影像数据。其中，45例患者为野生型，9例患者有突变。使用基于Matlab的IBEX软件包提取肿瘤图像的纹理特征。为了生成类平衡数据并提高预测稳定性，我们在交叉验证期间对突变组进行了数据增强。使用随机森林分类算法构建预测突变状态的模型，并采用留一法交叉验证进行评估。预测突变状态的随机森林模型的准确率为0.83，灵敏度为0.72，特异性为0.87，阳性预测值为0.65，曲线下面积为0.77，F值为0.68。CT影像组学是预测ccRCC患者突变状态的一种潜在可行方法。
