Enhancing Survival Prediction: The Potential of Whole-Brain Radiomics in Multimodal Neuroimaging

Abstract

Radiomics shows promise in enhancing predictions of overall survival (OS) and progression-free survival (PFS) in patients with glial brain tumors. The prognostic significance of imaging biomarkers derived from a whole-brain mask is still unclear. This study aimed to evaluate the potential of radiomics for predictive modeling of OS and PFS in patients with brain gliomas. We compared 13 prognostic models designed to predict OS and PFS, using clinical features alone, radiological biomarkers alone, and a combination of both. Our approach achieved C-index values of 0.900 for OS and 0.903 for PFS. Models built solely on imaging biomarkers exhibited the highest quality, whereas those based only on clinical signs showed the lowest quality. Given the limited data, it is unclear how reproducible the whole-brain radiomic features and corresponding models will be with new data. Nonetheless, there are reasons to view whole-brain radiomics as a promising avenue for further research.