RATIONALE AND OBJECTIVES: This study aims to establish a nomogram predictive model capable of identifying high-grade tumors among IDH-mutant astrocytomas exhibiting positive T2-FLAIR mismatch sign (T2FM) before surgery. MATERIALS AND METHODS: We collected T2FM-positive IDH-mutant astrocytomas from three distinct centers. Using data from one center as the training set (154 cases) to establish predictive model. Data from the remaining two centers served as external validation set (29 cases) to evaluate the model's performance. The assessment of the predictive model included the receiver operating characteristic (ROC) curve, calibration curve, decision curve analysis, area under the curve (AUC), sensitivity, and specificity. RESULTS: In the clinical features, we identified female sex (p=0.02) and the presence of enhancement within the tumor in contrast-enhanced T1-weighted imaging (p<0.001) as independent predictors of high-grade T2FM-positive IDH-mutant astrocytomas. We also found that radiomics features, such as LeastAxisLength, could aid in distinguishing WHO grades. Based on clinical and radiomics features, we developed an combined predictive model, which demonstrated superior performance compared to models relying solely on clinical or radiomics features. The combined predictive model achieved AUC of 0.819 and 0.858, sensitivity of 0.615 and 0.500, and specificity of 0.941 and 1.000, in the training and validation sets, respectively. CONCLUSION: We developed a predictive model based on gender, contrast enhancement, and radiomics features to predict high-grade tumors among T2FM-positive IDH-mutant astrocytomas with high specificity but low sensitivity.
He et al. (Tue,) studied this question.