An anthropomorphic phantom for advanced image processing of realistic 18F-FDG PET-CT oncological studies

In recent years, great efforts have been devoted to develop advanced image processing methods for oncological Positron Emission Tomography/Computed Tomography (PET-CT) images (e.g. segmentation, quantification, texture analysis) able to extract from images tumor characteristics hidden at naked eye, e.g. intra-tumor phenotypic heterogeneity. In order to accurately estimate such imaging descriptors, image processing methods need to be validated on datasets closer to the real clinical conditions, e.g. including lesions described by functional signal of irregular spatial distribution and heterogeneous intensity. The aim of this work was to realize an experimental dataset of a torso anthropomorphic phantom suitable for the assessment of advanced image processing in
¹⁸
F-fluorodeoxyglucose (
¹⁸
F-FDG) PET-CT oncological studies. The dataset consists into several
¹⁸
F-FDG PET-CT measurements of the RSD Alderson Thorax phantom including 42 lesions of irregular shape, different volumes (0.8-11.3 cc), and heterogeneous 18F-FDG uptake (actual lesion-to-background ratio (L/B
_GS
) 3-40). Simulation of real oncological lesions was obtained deriving the shape of the lesions from 18F-FDG PET-CT images of real patient lesions and printing 3D shells by a 3D printer. Radioactive gels consisting in a fast-setting alginate powder at different concentration allowed the simulation of heterogeneous uptake within the shells. The lesions were inserted in the RSD Alderson Thorax phantom, in the thorax and breast compartments, and measured by a current generation PET-CT system. Proofs of concept of the usefulness of the image dataset were provided. Such phantom image set could be downloaded from our and used by researchers for validation purposes.