<p>Glioblastoma multiforme (GBM) is the most aggressive brain tumor with a median survival under 15 months, making accurate survival prediction critical for treatment planning. Current prognostic models suffer from poor feature selection and limited predictive accuracy. We developed Time2Event, a machine learning framework that combines MRI-based radiomic analysis with particle swarm optimization (PSO) for enhanced survival prediction in GBM patients. Using T1-weighted contrast-enhanced and T2-FLAIR MRI sequences from 119 GBM patients, we extracted 412 radiomic features capturing tumor heterogeneity, shape, and intensity patterns. Unlike traditional filter-based methods, we employed PSO as a trainable metaheuristic optimizer to identify the most prognostically relevant features. These selected features were used to train four survival models: random survival forest (RSF), support vector regression (SVR), Cox proportional hazards (CPH), and gradient boosting survival analysis (GBSA). The PSO-optimized RSF achieved a concordance index of 91.66%, significantly outperforming conventional approaches (83.33% without PSO) and surpassing recent studies that achieved 75–85% accuracy even with additional genomic data. Texture-based features from both MRI sequences emerged as dominant predictors, reflecting tumor heterogeneity’s prognostic importance. This framework provides clinicians with an accurate, accessible risk stratification tool using routine MRI without requiring molecular markers. The substantial performance improvement demonstrates that intelligent feature optimization can unlock critical prognostic patterns in medical imaging, offering immediate clinical applicability for personalized GBM management and treatment decisions.</p>