Radiomics Outperforms Conventional Tumor Volume in Prognostic Stratification Across Independent Glioma Cohorts

Area of Science:

• Oncology
• Radiology
• Medical Imaging
• Machine Learning

Background:

• Tumor volume on MRI is a standard measure of glioma burden.
• Conventional volume may not reflect glioma's complex spatial and phenotypic characteristics.
• Need for advanced imaging analysis to improve prognostic accuracy.

Purpose of the Study:

• To compare the prognostic performance of segmentation-derived tumor volume versus MRI-derived radiomic risk scores in glioma.
• To evaluate the ability of radiomics to capture information beyond simple tumor burden.

Main Methods:

• Retrospective analysis of 1,447 patients across four independent glioma MRI cohorts.
• Discovery cohort (MU-Glioma-Post) for volumetric progression-free survival (PFS) analysis.
• External validation cohorts (UCSD-PTGBM, UCSF-PDGM, UPenn-GBM) for volume and radiomic score assessment.
• Head-to-head comparison of tumor volume and radiomic score prognostic performance.

Main Results:

• Tumor volume showed inconsistent prognostic performance across external cohorts.
• In the UPenn-GBM cohort, tumor volume was not associated with overall survival (OS).
• The radiomic score was independently associated with worse OS in UPenn-GBM (HR 1.634, p < 0.001).
• In head-to-head comparisons, the radiomic score remained significant while tumor volume did not.

Conclusions:

• Conventional tumor volume has limited and inconsistent prognostic value in glioma.
• MRI-derived radiomics demonstrate superior and reproducible prognostic performance.
• Radiomic phenotyping provides clinically relevant information beyond tumor burden measurements.