<![CDATA[Background: Radiomics has emerged as a promising quantitative approach in PET molecular imaging; however, its clinical translation remains limited by concerns regarding feature reproducibility under varying technical conditions. This systematic literature review aimed to identify and synthesize the main technical factors influencing the reproducibility of PET radiomic biomarkers. Methods: A systematic literature Review was conducted using the Scopus database with the search strategy TITLE-ABS-KEY (radiomics AND (PET OR PET/CT) AND (reproducibility OR "test-retest")). Fifteen eligible studies were included and narratively synthesized according to major technical domains: segmentation, motion and acquisition/reconstruction, resampling and discretization, scanner variability and harmonization, and phantom-based quality assurance. Results: The findings consistently indicate that segmentation strategy, respiratory motion, reconstruction parameters, voxel resampling, intensity discretization, and inter-scanner variability are major determinants of PET radiomic feature variability. Shape-based features and selected texture features, particularly gray-level co-occurrence matrix (GLCM) features, generally demonstrated higher repeatability than intensity-based and more complex texture features. While harmonization techniques, including AI-based approaches, and standardized phantom-based quality assurance protocols show promise for improving reproducibility, their generalizability remains insufficiently validated. Conclusions: PET radiomic biomarker reproducibility is highly sensitive to methodological variability across the imaging pipeline. Standardized acquisition protocols, robust feature selection, harmonization strategies, and routine phantom-based quality assurance are essential prerequisites before reliable clinical implementation. Further multicenter validation studies are needed to establish reproducible and clinically transferable PET radiomic biomarkers.]]>
Copyrights © 2026
