Three-Class Radiomics Models for IDH-Mutation and 1p19q-Codeletion Status Prediction in Adult-type Diffuse Gliomas

Motivation: Glioma patients with different IDH mutation and 1p19q-codeletion status indicate different treatment responses and prognosis. Goal(s): To predict the genetic subtypes of adult-type diffuse gliomas with three-class MRI radiomics. Approach: MRI based three-class radiomics models were used to predict the IDH mutation and 1p19q-codeletion status. Results: Support vector machine (SVM) classifiers combined with conventional MR sequences based on tumor ROI was proved to be the best performing diagnostic model. The model achieved area under the curve (AUC) of 0.963, 0.964, 0.925 for IDHwt, IDHmut-intact and IDHmut-codel prediction, with overall accuracy of 0.907. The model was validated both in internal and external validation set. Impact: Three-class MRI radiomics can preoperatively predict IDH and 1p19q-codeletion with satisfied performance, which is helpful for glioma risk stratification.