Fig. 2From: Preselection of robust radiomic features does not improve outcome modelling in non-small cell lung cancer based on clinical routine FDG-PET imagingRobustness studies. The influence of delineation variability or image was investigated using either a pair of scans for each patient as it is the case for the PET/CT/PET/MR and motion comparison, or different methods of defining the regions of interest (ROI), as applied for the delineation study. Radiomics calculation was performed on each individual ROI. Robust features were determined using the intraclass correlation coefficient (ICC)Back to article page