基于MRI影像组学列线图术前预测宫颈癌患者淋巴结转移的价值

Preoperative prediction of lymph node metastasis in cervical cancer patients based on MRI imaging omics column chart

目的 探讨MRI影像组学评分结合血液学参数的列线图模型对术前宫颈癌患者淋巴结转移的预测价值。方法 选取189例经手术病理证实的宫颈癌患者,以7:3的比例随机分为训练集132例和验证集57例。均术前接受MRI检查,于矢状位T2WI和对比增强T1WI沿肿瘤边缘逐层勾画三维感兴趣区域(VOI),并提取影像组学特征。通过最小绝对收缩与选择算子(LASSO)回归分析对提取的纹理特征进行初步筛选,并利用支持向量机(SVM)构建影像组学模型。同时,采用单-多因素逻辑回归构建影像组学列线图。使用ROC曲线来评估影像组学模型、临床模型和影像组学列线图的预测能力,并通过决策曲线分析(DCA)来评估影像组学列线图在临床应用中的价值。结果 影像组学列线图模型诊断性能最佳,优于临床模型、影像组学模型及血液学模型。校准曲线和Hosmer-Lemeshow检验结果提示列线图预测宫颈癌淋巴结转移与实际结果间一致性良好。结论 基于MRI影像组学列线图可高效准确的预测宫颈癌患者淋巴结转移情况,有较高应用价值。

Objective To explore the predictive value of the nomogram model of MRI imaging omics score combined with he⁃matological parameters for lymph node metastasis in preoperative cervical cancer patients.Methods 189 patients with cervical cancer confirmed by surgical pathology were selected and randomly divided into training set(132 cases)and validation set(57 cases)at a ratio of 7:3.All selected candidates underwent MRI preoperative examination,delineated three-dimensional regions of interest along the edge of the tumor in sagittal T2WI andcontrast-enhanced T1WI,and imaging omics features were extracted.The extracted texture features were initially selected by minimum absolute shrinkage and selection operator regression analysis,and the imaging omics model was constructed using the support vector machine.Meanwhile,single-multifactor logistic regression was used to construct the imaging omics nomogram.ROC curves were used to evaluate the predictive ability of imaging omics model,clinical model and imaging omics nomogram,and to evaluate the value of imaging omics nomogram in clinical application by decision curve analysis.Results The imaging nomogram model had the best diagnostic performance than the clinical model,imaging model and hematology model.The results of calibration curve and Hosmer-Lemeshow test suggested good agreement be⁃tween cervical lymph node metastasis and the actual results.Conclusion Based on MRI imaging omics nomogram can effi⁃ciently and accurately predict the lymph node metastasis in cervical cancer patients,which has high application value.