MRI影像组学预测肝外胆管癌腹部淋巴结转移价值研究

Investigating the Value of MRI Radiomics for Predicting Abdominal Lymph Nodes Metastasis in Patients with Extrahepatic Cholangiocarcinoma.

目的探讨MRI影像组学术前预测肝外胆管癌患者淋巴结转移的价值。方法回顾性分析63例肝外胆管癌患者的MRI图像,按8∶2随机将淋巴结分为训练组(n=65)及验证组(n=16)。利用飞利浦ISD软件于轴位DWI图像手动勾画局部淋巴结的三维感兴趣区,并复制到对应的ADC图上,利用RyRadiomics软件包提取影像组学特征。采用皮尔逊相关系数及Kruskal-Wallis算法筛选出30种最佳影像组学特征,使用支持向量机建立影像组学模型。利用受试者工作特征曲线评价模型预测效能。结果良、恶性淋巴结的短径(P=0.009)、长径(P=0.002),差异有统计学意义;最终14种影像组学特征参与建立MRI影像组学模型,影像组学模型在训练集及测试集中曲线下面积分别为0.895及0.917。结论基于淋巴结本身的MRI影像组学模型能较好地预测肝外胆管癌患者的淋巴结转移,有助于指导临床选择精准淋巴结清扫范围并改善患者预后。

Objective To investigate radiomics value based on MRI in predicting lymph node metastasis in patients with extrahepatic cholangiocarcinoma before operation.Methods MRI images of 63 patients with extrahepatic cholangiocarcinoma were retrospectively analyzed.All lymph nodes were randomly divided into training group(n=65)and validation group(n=16)on a ratio of 8∶2.The lymph nodes 3D regions of interest were manually delineated on the axial diffusion weighted imaging(DWI)using Philips ISD software and copied to the corresponding apparent diffusion coefficient(ADC)image.Radiomics features of lymph nodes were extracted using the RyRadiomics package.The 30 optimal radiomics features were selected using Pearson correlation coefficient(PCC)and the Kruskal-Wallis algorithms.A support vector machine(SVM)was used to build the radiomics model.The diagnostic effciency of the radiomics model was evaluated using receiver operating characteristic curve(ROC).Results There were statistically differences in max diameter(P=0.002)and min diameter(P=0.009)between benign lymph nodes and maligant lymph nodes.The 14 radiomics features finally used to build the radiomics model.The AUC of radiomics model was 0.895 in training group and 0.917 in validation group.Conclusion MRI radiomics model based on the lymph nodes could accurately predict lymph node metastasis in patients with extrahepatic cholangiocarcinoma,which could help guide clinical selection of precise lymph node dissection area and improve prognosis of patients.