Introduction:
Gliomas were highly aggressive primary brain tumors that require precise grading for effective treatment planning. Current grading relies on histopathological analysis, which can be invasive and subjective. Radiomics,a non-invasive approach that extracts quantitative features from medical images,may offer enhanced diagnostic accuracy.

Material and methods:
This retrospective study analyzed 263 glioma patients from January 2021 to December 2023. Patients were categorized into low-grade glioma (LGG) and high-grade glioma (HGG) groups based on WHO classifications. MRI images were used to extract radiomic features,compared to traditional imaging characteristics. PyRadiomics facilitated extraction of features such as tumor sphericity and GLCM contrast.Logistic regression with 10-fold cross-validation assessed the diagnostic accuracy,with Receiver Operating Characteristic (ROC) curve analysis and the DeLong test (Bonferroni-adjusted) used to compare the predictive performance of the models.

Results:
Significant differences were observed in radiomics features between LGG and HGG groups, including tumor sphericity and GLCM contrast,with higher heterogeneity in HGG.The radiomics-based model achieved a bootstrap-corrected area under the curve (AUC) of 0.847 (95% CI: 0.80-0.89), sensitivity of 82.0%, and specificity of 38.7%.Traditional imaging characteristics alone demonstrated an AUC of 0.806 (95% CI: 0.76-0.85), with sensitivity and specificity at 80.0% and 35.4%, respectively. The combined model of radiomics and traditional imaging provided an enhanced AUC of 0.872 (95% CI: 0.83-0.91), with improved accuracy (80.4%), sensitivity (84.0%), and specificity (44.8%).

Conclusions:
Radiomics-based models significantly enhance the diagnostic grading of gliomas compared to traditional imaging alone by capturing subtle tumor heterogeneity and complexity. The integration of radiomics and traditional imaging offers superior diagnostic performance, potentially guiding more accurate treatment strategies.