基于常规MRI的影像组学预测脑胶质瘤IDH突变的效果分析

目的探讨基于常规MRI的影像组学分析预测脑胶质瘤IDH突变的效果。方法回顾性分析经术后病理证实的172例脑胶质瘤患者(82例突变型和90例野生型)的临床资料。所有患者术前均行增强MRI检查,术后病理对IDH进行分子检测。由2名神经影像医师在FLAIR和T1加权对比增强(CE-T1WI)序列上逐层勾画肿瘤实体及周围水肿病灶,采用组内相关系数(ICC)评价2名医师分别勾画的感兴趣区域组学特征一致性,然后提取影像特征。分别使用方差阈值法、单变量特征选择法、最小绝对收缩和选择算子(LASSO)对组学特征进行筛选和降维。采用χ²检验或Fisher精确检验评价2组胶质瘤影像学特征的统计学差异。采用五折交叉验证法的分类方式,通过Logistic回归分析建立放射组学模型,采用诺莫图展示模型结果,采用校准曲线验证模型的可靠性。结果在CE-T1WI序列筛选出13个组学特征,在FLAIR序列筛选出7个组学特征。癫痫、WHO分级、影像所见强化和水肿对于IDH突变状态预测具有统计学意义(P<0.05)。影像组学模型在训练集中的AUC为0.833(95%CI0.778~0.889),在测试集中的AUC为0.753(95%CI0.628~0.895)。结论影像组学在预测脑胶质瘤IDH突变方面具有较好的性能。

Effect of radiomics based on conventional MRI in predicting IDH mutation in glioma

Objective To evaluate the effect of radiomics based on conventional MRI in predicting IDH mutation in glioma.MethodsThe clinical data of 172 glioma patients(82 cases of IDH-mutant type and 90 cases of IDH-wild type)confirmed by postoperative pathology were analyzed retrospectively.All patients underwent preoperative enhanced MRI examination and postoperative pathology for molecular detection of IDH.The tumor entity and the surrounding edema lesions were delineated layer by layer by two neuroimaging physicians on FLAIR and T1-weighted contrast-enhanced(CE-T1WI)sequences,and the intra-group correlation coefficient(ICC)was used to evaluate the consistency of radiomics features of the regions of interest delineated by the two physicians respectively,and then the image features were extracted.The variance threshold method,univariate feature selection,and least absolute shrinkage and selection operator(LASSO)were used to screen and reduce the dimension of the radiomics features,respectively.Theχ² test or Fisher exact test was used to evaluate the statistical differences in the imaging features of glioma between the two groups.The radiomics model was established by Logistic regression analysis using the classification method of five-fold cross-validation,the results of the model were displayed by Nomogram,and the reliability of the model was verified by calibration curve.Results A total of 13 radiomics features were screened from the CE-T1WI sequence and 7 radiomics features were screened from the FLAIR sequence.Epilepsy,WHO grade,image enhancement and edema had statistically significance for IDH mutation status prediction(P<0.05).The area under the curve(AUC)of the radiomics model was 0.833(95%CI:0.778 to 0.889)in the training set and 0.753(95%CI:0.628 to 0.895)in the test set.Conclusion Radiomics has good performance in predicting the IDH mutation in glioma.