Tumor microstructural biomarkers from time-dependent diffusion MRI for predicting IDH mutation status in...

To explore the value of microstructural parameters derived from time-dependent diffusion MRI (TDD-MRI) in predicting isocitrate dehydrogenase (IDH) mutation status in non-enhancing gliomas.

A total of 112 patients with non-enhancing gliomas, comprising 90 IDH-mutant and 22 IDH-wildtype gliomas, were enrolled retrospectively. Seven microstructural features were calculated from TDD-MRI. After selecting features via least absolute shrinkage and selection operator regression and Boruta algorithm, Firth's logistic regression was applied to develop three predictive models, including quantitative model, clinical model, and integrative model. Model performance was evaluated using area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, positive predictive value, and negative predictive value. A 5-fold cross-validation and bootstrap were conducted to provide a stable estimate of model performance.

Among all microstructural parameters, cellularity exhibited the most favorable diagnostic efficacy, achieving an AUC of 0.758 and an accuracy of 0.830. The quantitative model outperformed single microstructural parameters with an AUC of 0.828 and an accuracy of 0.857, whereas the clinical model exhibited an AUC of 0.864 and an accuracy of 0.866. The integrative model exhibited the highest diagnostic efficacy, with an AUC of 0.903 and an accuracy of 0.884. For stability, the quantitative, clinical and integrative models yielded 5-fold cross-validation AUCs of 0.836, 0.883 and 0.908, and bootstrap AUCs of 0.821, 0.862 and 0.893, respectively.

TDD-MRI-derived cellularity is the most potent microstructural predictor of IDH status in non-enhancing gliomas. The integrative model achieved superior diagnostic performance, providing a robust and clinically translatable framework for non-invasive glioma genotyping.