Glioblastoma radiomics study questions feature robustness for survival prediction

A new study published on arXiv investigated the relationship between the robustness of radiomic features and their predictive utility in glioblastoma (GBM) survival modeling. Researchers analyzed 4,752 radiomic features from multiparametric MRI scans of GBM patients, assessing feature robustness using intraclass correlation coefficients. The findings indicated that a significant portion of features were not robust, and filtering for robustness did not improve survival prediction models. The study suggests that robustness alone may not be a sufficient criterion for selecting features in radiomics-based survival analysis. AI

IMPACT This research highlights potential limitations in using radiomic features for glioblastoma survival prediction, suggesting a need for refined feature selection methods.

RANK_REASON Academic paper detailing research findings. [lever_c_demoted from research: ic=1 ai=0.4]

Read on arXiv cs.CV →

• Coxnet
• glioblastoma
• Gradient Boosting Survival Analysis
• magnetic resonance imaging
• radiomics
• random survival forest
• University of Pennsylvania Glioblastoma Imaging, Genomics, and Radiomics
• UPENN-GBM

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →