基于术前MRI影像组学及临床特征的卵巢癌淋巴结转移预测模型构建及验证

Establishment and validation of predictive model for lymph node metastasis of ovarian cancer based on preoperative MRI imaging featuresand clinical characteristics

目的探讨卵巢癌淋巴结转移的危险因素,并构建基于术前MRI影像组学及临床特征的卵巢癌淋巴结转移的列线图。方法选取2018年2月—2021年7月期间于安徽医科大学第一附属医院就诊的225例卵巢癌患者作为研究对象,根据淋巴结转移情况分为淋巴结转移组和无淋巴结转移组。应用Logistic回归分析筛选卵巢癌淋巴结转移的危险因素。利用LIFEx软件提取卵巢癌患者手术前的核磁共振(MRI)影像组学特征,将所选MRI影像组学与其对应非零系数乘积的线性组合确定其MRI影像组学评分。建立MRI影像组学、临床特征模型。通过整合优化的MRI影像组学及临床特征模型,采用R(R4.2.0)软件建立卵巢癌淋巴结转移的列线图,并对模型进行内部验证。结果225例卵巢癌中有95例患者出现淋巴结转移,淋巴结转移的发生率为42.22%(95/225)。Logsitc回归分析结果显示,临床分期、病灶位置、分化程度及糖尿病等是卵巢癌淋巴结转移的危险因素(P<0.05)。结合多因素结果及MRI影像组学参数提取构建了3组预测模型,其中包括临床模型1组,MRI影像组学模型1组,组合模型(临床-MRI影像组学模型)1组,其中临床-MRI影像组学模型的AUC(0.862)最高。构建的临床-MRI影像组学关于卵巢癌淋巴结转移的列线图模型的结果显示,校正曲线显示预测值和实际值具有良好的拟合度,模型的ROC曲线下面积为0.862(95%CI:0.790~0.934),决策曲线显示阈值概率为22%~80%时,列线图预测卵巢癌淋巴结转移的净获益值较高。结论基于术前MRI影像组学及临床特征的卵巢癌淋巴结转移的列线图模型具有较高的准确率和较好的临床应用价值,能够用于术前卵巢癌淋巴结转移的识别。

Objective To investigate the risk factors of lymph node metastasis of ovarian cancer,and to construct a nomogram of lymph node metastasis of ovarian cancer based on preoperative MRI imaging featuresand clinical characteristics.Methods 225 patients with ovarian cancer who came to our hospital from February 2018 to July 2021 were selected as research objects,and were divided into the lymph node metastasis group and the non-lymph node metastasis group according to the lymph node metastasis.Logistic regression analysis was used to screen the risk factors of lymph node metastasis of ovarian cancer.The MRI imaging features of ovarian cancer patients before surgery were extracted by LIFEx software,and the linear combination of these selected MRI imaging featuresand their corresponding non-zero coefficientswas used to determine the MRI imaging score.To establish the model of MRI imaging features and clinical characteristics.By integrating the optimized modal of MRI imaging features and clinical characteristics,R(R4.2.0)software was used to establish the nomogram of ovarian cancer lymph node metastasis,and the model was internally validated.Results Among 225 cases of ovarian cancer,95 patients had lymph node metastasis,and the incidence of lymph node metastasis was 42.22%(95/225).Logsitc regression analysis showed that clinical stage,focus location,differentiation degree and diabetes were risk factors for lymph node metastasis of ovarian cancer(P<0.05).Combined with the results of multiple factors and the extraction of MRI imageomics parameters,three prediction models were constructed,including one clinical model,one MRI imageomics model,and one combined model(clinical MRI imageomics model).Among the three prediction models,it was found that the area under curve(AUC)(0.862)of the clinical MRI imageomics model was the highest.The results of the nomogram model of clinical MRI imaging features on ovarian cancer lymph node metastasis showed that the correction curve showed that the predicted value had a good fit with the actual value;The area under the ROC curve of the model was 0.862(95%CI:0.790-0.934);The decision curve showed that when the threshold probability was 22%-80%,the net benefit value of nomogram in predicting lymph node metastasis of ovarian cancer was high.Conclusion The nomograph model of lymph node metastasis of ovarian cancer based on preoperative MRI imaging featuresand clinical characteristics has a high accuracy and good clinical application value,and can be used for preoperative identification of lymph node metastasis of ovarian cancer.