影像组学联合表观弥散系数评估低级别脑胶质瘤IDH-1基因突变状态

Evaluate the status of IDH-1 gene mutation by the combine model constructed by apparent diffusion coefficient and radiomics

目的探讨影像组学联合表观弥散系数(ADC)评估低级别脑胶质瘤异柠檬酸脱氢酶-1(IDH-1)基因突变状态。方法回顾性纳入荆门市第一人民医院2015-05—2020-12入院且术后经病理证实为低级别脑胶质瘤患者。基于患者术前磁共振弥散加权成像转为表观弥散系数图,分别计算患者病灶的表观弥散系数以及影像组学参数。将IDH-1野生型患者与突变型患者按照7∶3比例分为训练组和测试组,基于训练组患者按照IDH-1是否突变为研究目的,对影像组学参数进行特征降维,构建影像组学标签(Rad_score),最后联合病灶ADC值构建影像联合模型评估IDH-1是否突变,并采用Hosmer-Lemeshow以及临床决策线分析联合模型的临床价值。结果基于训练组中所有患者的影像组学特征以IDH-1是否突变为分类标签采用LASSO进行影像组学特征降维,取最小惩罚系数Logλ=0.0052对应的特征集合,共15个影像组学特征构建影像组学标签Rad_score。训练组中以及测试组中IDH-1野生型患者Rad_score低于IDH-1突变型患者,且差异有统计学意义(P=0.0000025 vs 0.0018)。联合训练组患者病灶ADC和Rad_score以IDH-1是否突变为研究目的构建多元逻辑回归模型Imagemodel,训练组中Imagemodel的AUC值>ADC>Rad_score(0.98 vs 0.95 vs 0.83),测试组中Imagemodel的AUC值>ADC>Rad_score(0.99 vs 0.97 vs 0.83)。结论ADC联合影像组学构建影像联合模型可协助临床术前评估低级别脑胶质瘤患者IDH-1的突变状态。

Objective To explore the feasibility of preoperative evaluation of glioma isocitrate dehydrogenase-1(IDH-1)gene mutation by imaging combined with apparent diffusion coefficient.Methods Patients admitted to our hospital from May 2015 to De⁃cember 2020 who were pathologically confirmed to be low-grade gliomas after surgery were retrospectively enrolled.Patients were en⁃rolled strictly according to the inclusion and exclusion criteria.The apparent dispersion coefficient and imaging omics parameters of the patients were calculated based on the preoperative DWI of the patients,respectively.The mutant and wild-type IDH-1 patients ac⁃cording to the proportion and randomly divided into exercise group and test group,based on the training group patients according to whether mutations IDH-1 for research purposes,the parameters of weft,image omics tag to build image,finally combined ADC build images combined model to evaluate whether mutations,IDH-1 and the Hosmer-Lemeshow and analysis the clinical value of combined model of clinical decision line.Results Based on the imaging omics characteristics of all patients in the training group,the image omics label Rad_score was constructed by taking IDH-1 mutation as classification label and using LASSO to reduce the dimension of the imaging omics features.The feature set corresponding to the minimum penalty coefficient Logλ=0.0052 was selected,and a total of 15 imaging omics features were used to construct the image omics label Rad_score.The Rad_score of IDH-1 wild-type patients in the training group and the test group was lower than that of IDH-1 mutant patients,and the difference was statistically significant(P=0.0000025 vs 0.0018).In the training group,the AUC value of Imagemodel was higher than that of ADC and Rad_score(0.98 vs 0.95 vs 0.83),while in the test group,the AUC value of Imagemodel was higher than that of ADC and Rad_score(0.99 vs 0.97 vs0.83).Conclusion The combined imaging model of ADC combined with imaging omics can assist in the preoperative evaluation of IDH-1 mutation status in patients with low-grade glioma.