Results: Radiomics classifiers achieved AUCs of 0.81-0.94for pseudoprogression, improving with molecular data.AI-assisted assessment increased inter-observer agreement (CCC 0.770.91)and outperformed less-experienced readers.Automated volumetric MRI analysis showed moderate agreement with radiologist assessments (F1 0.59-0.76)and could stratify patients by survival.Machine-and deeplearning models also predicted [' and molecular characteristics with 80-90% accuracy.These results demonstrate that AI enhances reproducibility, diagnostic accuracy, and prognostic assessment in glioma management.Conclusions: AI holds much promise for its role in revolutionizing the management of glioma by providing objective, reproducible, non-invasive treatment response and patient prognosis assessment.Current models show high accuracy in predicting pseudoprogression, molecular markers, and overall survival, challenges such as data heterogeneity, algorithm interpretability, and need for external validation remain.
Rajput et al. (Sun,) studied this question.