LAUSR

PurposeTo perform an updated review regarding the influence of methodological conditions on radiomics analyses of 18F-FDG PET imaging.MethodsWe performed a systematic review of the literature using PubMed/Medline and Google scholars, with multiple research keywords for each organ accompanying the terms “radiomics”, “texture”, “heterogeneity”, “FDG”, “PET” and “PET/CT”. The review considers methodological studies for shape, histogram and textural features extracted from FDG PET imaging. The references cited in the retrieved articles were also explored to find additional studies. The search was limited to English language. Preclinical and animal studies were not included in the review. A total of 44 original articles were considered for the review.ResultsIn the same conditions, repeatability is very variable among the radiomics features in FDG PET. The vast majority of features are sensitive to acquisition and reconstruction settings independently of the category or order of features. Similar sensitivity and variability are observed with respect to respiratory motion, pre-processing, segmentation method, and specifically for textural features, discretisation of grey levels and implementation/parameterisation of texture matrices.ConclusionRadiomics features are sensitive to almost all factors involved in the PET/CT workflow, from the generation of the images, to the radiomics implementation choices. This strongly supports the need for standardization efforts of these conditions to enable implementation in clinical routine and multicentric studies. However, to date the repercussion of these features fluctuations on the clinical endpoints has been rarely studied. Similar standardization and consensus will also be needed for the statistical analysis and machine learning aspects involved in radiomics analyses of FDG PET images.