RadiomiK: a phantom to test repeatability and reproducibility of Radiomic Features extracted from CT images

Purpose/Objective Radiomics has been demonstrated to have a role in several clinical processes. Although Radiomic approach is interesting it suffers from several noise sources, associated with image acquisition and post-processing (1). Understanding noise sources due to image acquisition permits to guide the creation of local prospective imaging protocols. Aim of this work is to present a phantom developed to test reproducibility and repeatability of radiomic features extracted from CT images. Material/Methods The phantom- RadiomiK (fig. 1a) comprises 23 elements imbedded in a layer of epoxy resin. The shape, materials and filling textures were chosen to produce a wide range of radiomics feature values capable to mimic those found in CT images of human being. 15 elements with cubic, cylindrical or conical shape, (1cm side, diameter and high) were fabricated with a 3D printer using PLA, FLEX and PETG. Honeycomb and gyroid textures, with air-filled holes and different filling percentage were employed. 3 cubes and 2 cylinders were made using slabs of Solid Water, Cortical Bone and Lung (Gammex-RMI, Middleton, WI, USA). Finally, 3 elements made with 25 mini-cubes (2mm side) were assembled creating three different patterns. Tests- 9 CT studies were acquired with a Somatom Definition Flash (Siemens, Erlangen, Germany) in helical mode, using an iterative reconstruction kernel i50f with strength 3, 2mm slice thickness, 0,25 pixel size, 120KV, 300, 250, 200, 175, 150, 125, 100, 75, 50 mAs. MIM-Maestro (MIM Software, Cleveland, OH) was used to manually segment 4 ROIs and the Moddicom package (1) to extract Radiomic features. Results A CT cross section of Radiomik (fig 1b) shows the different materials, textures and fill levels (20%, 60%, 100%) of the 23 elements. The mean HU values ranged between -630 HU to 1420HU. Epoxy resin HU was 90. Contrast between inserts and background is good but the boundary of inserts is not sufficiently sharp to enable automatic segmentation with threshold or region growing algorithms. The 4 ROIs considered for Radiomic Features extraction are highlighted in fig 1b). For each radiomic feature (n=74), we calculated the Spearman's correlation coefficient between feature value and acquisition mAs (fig.2). Radiomic features showing the highest dependency from mAs were mostly not shared across the different ROIs, suggesting a not-negligible interplay between ROI characteristics and CT protocol setting in feature value estimation. Conclusions The performed analysis focused on investigating the usability and usefulness of the proposed phantom, and it showed that Radiomik can be an useful tool to study Radiomic Feature reproducibility and repeatability.