Integrating Radiological Analysis and Spatiotemporal Modeling to Uncover Prognostic Heterogeneity in Multifocal Glioblastoma

Multifocal glioblastoma (mGB) is an aggressive brain tumor characterized by clinical and molecular heterogeneity. We reconstructed three-dimensional tumor structures using magnetic resonance images from 35 patients with mGB (wanglab.shinyapps.io/mfgExplorer) and identified two subgroups based on the size of the second‑largest lesion (L2), with larger L2 lesions associated with poorer survival. L2‑large mGBs exhibited frequent genetic alterations and transcriptomic signatures enriched for epithelial-mesenchymal transition and cell migration pathways. Computational modeling and evolutionary analyses indicated that earlier and more extensive tumor cell migration is associated with L2-large morphology and may contribute to poor prognostic outcomes. Our model also implied the widespread existence of radiographically invisible lesions, potentially explaining the inevitable recurrence of this disease, while the proposed mGB score revealed prognostic heterogeneity among mGB patients. Collectively, our work bridges radiological, molecular, and computational analyses to support improved risk stratification and provide a theoretical basis for understanding mGB progression.