Prediction Models of IDH and ATRX Gene Status in Diffuse Gliomas Based on Visually Accessible Rembrandt Images...

Identifying isocitrate dehydrogenase (IDH) mutation and α-thalassemia/mental retardation syndrome X-linked (ATRX) mutation status is helpful for diagnosis and specific classification of diffuse gliomas, while currently, the detection of IDH and ATRX status mainly relies on invasive methods. In this study, we aimed to predict IDH and ATRX mutation status of diffuse gliomas utilizing clinically available MRI Visually Accessible Rembrandt Images (VASARI) features.

Five hundred ninety-two patients (352 IDH wild-type and 240 IDH-mutant patients) with pathologically proved diffuse gliomas from our institution were randomly divided into training set (n=414) and validation set (n=178) for IDH mutation prediction according to a ratio of 7 to 3. Patients with IDH mutant were further stratified into ATRX mutant (n=109) and ATRX wild-type (n=131) subgroups, with the cohort then divided into training set (n=168) and validation set (n=72) for ATRX mutation prediction. Two radiologists independently analyzed the patients' MR images based on the VASARI feature set. Multivariable logistic regression analysis was employed to develop the prediction models. Receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA) were utilized to validate the models and nomograms were developed to visualize the models.

For IDH prediction, 6 VASARI features combined with age and relative ADC values contributed to the model, with the area under the curve (AUC) of 0.96 (0.94-0.98) in training set and 0.92 (0.88-0.97) in validation set. For ATRX prediction, 3 VASARI features combined with age and minimum ADC values contributed to the model, with the AUC of 0.76 (0.68-0.83) in training set and 0.71 (0.58-0.83) in validation set. The DCA and calibration plots further confirmed the clinical utility of the 2 nomograms for IDH and ATRX prediction.

The integration of MRI VASARI features and clinical data demonstrates strong predictive capability for IDH mutation status and moderate predictive capability for ATRX status in diffuse gliomas.