Predictive Role of the Apparent Diffusion Coefficient and MRI Morphologic Features on IDH Status in Patients With Diffuse Glioma: A Retrospective Cross-Sectional Study

摘要: Purpose To evaluate isocitrate dehydrogenase (IDH) status in clinically diagnosed grade II~IV glioma patients using the 2016 World Health Organization (WHO) classification based on MRI parameters. Materials and Methods One hundred seventy-six with confirmed WHO were retrospectively investigated as study set, including lower-grade (WHO II, n = 64; III, 38) glioblastoma IV, 74). The minimum apparent diffusion coefficient (ADCmin) tumor contralateral normal-appearing white matter (ADCn) rADC (ADCmin to ADCn ratio) defined calculated. Intraclass correlation (ICC) analysis was carried out interobserver intraobserver agreement for ADC measurements. Interobserver morphologic categories evaluated by Cohen's kappa analysis. nonparametric Kruskal-Wallis test used determine whether measurements subtypes related. By univariable analysis, if differences a variable significant (P 0.8; κ >0.6), then it chosen predictor variable. performance of area under receiver operating characteristic curve (AUC) several machine learning models, logistic regression, support vector machine, Naive Bayes Ensemble. Five evaluation indicators adopted compare models. optimal model developed final predict IDH 40 subsequent set. DeLong AUCs. Results In six measured variables (rADC, age, enhancement, calcification, hemorrhage, cystic change) selected model. Logistic regression had better than other Two predictive 1 (including all variables) 2 (excluding calcification), correctly classified an AUC 0.897 0.890, respectively. set performed equally well prediction, indicating effectiveness trained classifier. subgroup revealed that predicted LGG GBM accuracy 84.3% (AUC 0.873) 85.1% 0.862) 70.0% 0.762) 0.833) Conclusion Through use machine-learning algorithms, accurate prediction IDH-mutant versus IDH-wildtype achieved adult diffuse gliomas via noninvasive MR imaging characteristics, values features, which are considered widely available most clinical workstations.