Fig. 3From: Radiomics for glioblastoma survival analysis in pre-operative MRI: exploring feature robustness, class boundaries, and machine learning techniquesRobustness of radiomic features against perturbations, measured by the intraclass correlation coefficient (ICC(2,1)). The values shown are averages of the lower bound of the 95th percentile of the ICC(2,1) across all segmentation types, MR sequences and features within a given class. Only relevant results are shown, e.g. not showing noise impact for shape features, since no intensity information was used for this featureBack to article page