ABSTRACT
The conversion of images, such as those from positron emission tomography (PET) and x-ray computed tomography (CT), into mineable high-dimensional data is known as radiomics. Radiomics is motivated by the concept that biomedical images contain information that reflects underlying physiology and molecular biology, and that these relationships can be revealed via quantitative image analyses. Thus an understanding of quantitative PET/CT accuracy, and how variations in acquisition and image reconstruction parameters influence imaging features in the absence of true underlying biologic effects, is essential to PET/CT radiomics. In this chapter, we first provide an overview of quantitative PET/CT, with emphasis on uncertainties and their clinical impact on quantitative precision and accuracy. We highlight ongoing efforts towards PET/CT standardization and harmonization as part of quantitative imaging biomarker quality assurance programs. With a basis in quantitative PET/CT, we introduce PET/CT radiomics, list additional sources of radiomic data processing uncertainty, propose guidelines to improve transparent reporting of PET/CT radiomic features, and emphasize the growing need for robust omics data analytics to help ensure reproducibility and clinical translation of scientific findings. Lastly, we summarize future directions beyond PET/CT radiomics, including PET/MR radiomics, non-FDG-PET radiomics, normal tissue function PET radiomics, and deep learning of PET/CT images.
