Assessing radiomic feature robustness to interpolation in 18F-FDG PET imaging - data

A set of feature extraction results used to assess feature robustness to interpolation in 18F-FDG PET imaging. Morphological, first-order, and texture features (n=141), were extracted from segmented tumour volumes. Patients had biopsy-proven oesophageal cancer and had undergone PET/CT imaging as part of staging. Data is from 441 patients, split (80%/20%) into testing (n=353) and validation (n=88) dataset. Both linear and spline interpolation methods were explored. Features were extracted after interpolation to 6 different isotropic voxel sizes (1.5mm, 1.8mm, 2.0mm, 2.2mm, 2.5mm, 2.7mm). Each csv file relates to a specific interpolation method, dataset type (testing or validation), and voxel size.