Detecting isocitrate dehydrogenase mutation status in grade IV gliomas using stochastic differential equations

The status of the Isocitrate Dehydrogenase (IDH) mutation is crucial in determining the treatment plan for gliomas, particularly in grade IV tumors. The current gold standard for identifying this status relies on invasive procedures, such as biopsy or surgery, which are associated with risks and high costs. This study proposes an innovative, non-invasive method for identifying IDH mutation status in grade IV gliomas using conventional MRI images. This approach leverages a statistical modeling framework based on stochastic differential equations (SDE). Given that MRI images exhibit inherent stochastic and non-stationary properties, they can be effectively modeled as stochastic processes. In the proposed method, a fractional Laplacian operator is applied to MRI images, rendering the pixel intensities statistically independent and resulting in a zero-mean alpha-stable distribution. The parameters of this distribution, along with additional features, are used as input for four classifiers: Support Vector Machine, K-Nearest Neighbor, Random Forest, and Extreme Gradient Boosting, which automatically detect the IDH mutation status. The results revealed a classification accuracy of 99.29%, sensitivity of 99.06%, specificity of 99.52%, and an area under the receiver operating characteristic curve (AUC) of 100% for the first dataset, and an accuracy of 97.60%, sensitivity of 97.14%, specificity of 97.92%, and AUC of 99% for the second dataset in the slice-wise approach. On the other hand, the patient-wise approach achieved accuracies of 91.67% and 83.33% for the first and second datasets, respectively. These outcomes underscore the remarkable effectiveness of the proposed approach in detecting IDH mutation status in grade IV gliomas.