基于增强CT影像组学模型预测肺腺癌ALK突变

Radiomics Models Based on Enhanced CT for Predicting ALK Mutation in Lung Adenocarcinoma

目的探讨基于增强CT影像组学特征建立的模型对肺腺癌间变性淋巴瘤激酶(ALK)基因突变状态的预测价值。资料与方法回顾性分析经病理学检查证实的210例肺腺癌患者,在术前增强CT图像中勾画感兴趣区,再用A.K.软件提取影像组学特征并筛选出与ALK突变高度相关的特征,构建影像组学特征预测模型。联合具有独立预测效能的一般特征构建诺模图模型,并分别评价其预测效能。结果通过特征筛选后,最终选取19项影像组学特征用于建立最佳预测模型。该模型在训练集AUC值为0.89,敏感度和特异度分别为83%和82%;在验证集AUC值为0.79,敏感度和特异度分别为74%和85%。一般特征的加入未能显著提高影像组学特征模型的预测效能(P>0.05)。结论基于增强CT影像组学特征建立的模型对肺腺癌ALK基因突变状态具有较好的预测价值。

Purpose To explore the value of radiomics model based on enhanced CT in predicting the mutation state of anaplastic lymphoma kinase(ALK)gene in lung adenocarcinoma.Materials and Methods 210 patients with lung adenocarcinoma confirmed by surgical pathological examinations were retrospectively analyzed.The area of interest in their preoperative enhanced CT images were delineated,and then used A.K.software to extract quantitative radiomics features.Features highly related to ALK mutations were selected to construct radiomics prediction model.The radiomics features combined with independent general predictive factors to construct radiomics nomogram prediction model.The efficiency of radiomics prediction model and radiomics nomogram prediction model were evaluated separately.Results 19 radiomics features highly related to ALK mutations were finally selected to establish radiomics prediction model.In the training set,the AUC value of the model was 0.89,the sensitivity and specificity were 83%and 82%,respectively.In the verification set,the AUC value was 0.79,the sensitivity and specificity were 74%and 85%,respectively.The addition of general features did not significantly improve the prediction efficiency of the radiomics model(P>0.05).Conclusion The radiomics model based on enhanced CT is of good value in predicting the mutation state of ALK gene in lung adenocarcinoma.