This dataset contains the full set of CT images and segmentation masks generated for a systematic investigation of radiomic feature reproducibility and noise sensitivity using two biological phantoms composed respectively of bovine and swine tissue. The phantom was scanned under all combinations of four tube voltages (80–140 kV), four tube currents (50–300 mA), and four slice thicknesses (0.6–5 mm), yielding 128 test–retest acquisition pairs. The dataset includes: DICOM images for all acquisition settings Segmentations of volumes of interest Segmentation of background-air regions for noise assessment Cross-site validation data from a second phantom experiment This dataset supports research on radiomic reproducibility, robustness across acquisition variability, and methods for distinguishing biologically meaningful features from noise-driven pseudo-texture. It is intended as a resource for radiomics standardization, methodological benchmarking, and development of robust quantitative imaging biomarkers.