Radiomic Score as a Noninvasive Predictor of Overall Survival in Gliomas: A Multicenter Study

Motivation: Developing a prognostic score out of high-dimensional radiomics features could reflect the underlying pathophysiology of gliomas, enabling an easy-to-use tool for predicting patient prognosis. Goal(s): To develop a radiomics-based risk score (RS) for a preoperative, noninvasive assessment of overall survival (OS) in gliomas. Approach: T2w MRI of a local discovery dataset was used for developing RS, which was then externally validated on an independent dataset (UCSF-PDGM). Results: The tumor grade, age, and RS were significantly associated with OS (P<0.005, for all). RS showed significant prognostic performance in both discovery and validation datasets (P=1.48E-8 and P=1.84E-15, respectively). Impact: Risk stratification in gliomas based on RS could help identify high-risk patients, who may benefit from more aggressive treatment and closer follow-up, enabling personalized approach to care. The RS was established on preoperative T2w-MRI and stably predicts the glioma OS.