Prognostic value of radionics based on ¹⁸F-FDG PET/CT in Hodgkin lymphoma: A systematic review

To systematically evaluate the available evidence on the prognostic value of ¹⁸F-FDG PET/CT-based radiomics in Hodgkin lymphoma (HL), its methodological limitations, and potential clinical implications.

A systematic search was conducted in PubMed following PRISMA guidelines. Observational studies or radiomic analyses within clinical trials were included if they evaluated patients with HL and assessed the prognostic value of radiomic features extracted from PET/CT by analyzing their association with survival, treatment response, or treatment refractoriness. Reviews, meta-analyses, and studies without clinical outcomes were excluded. Study quality was assessed using the Radiomics Quality Score (RQS). Given the limited number of studies, a narrative synthesis was performed.

Of the 53 initially identified references, 9 retrospective studies met the inclusion criteria. In 7 studies, radiomic or multimodal models showed greater prognostic performance for progression-free survival and overall survival than clinical or volumetric models, with AUC values ranging from 0.86 to 0.92. Textural features derived from GLRLM, GLSZM, and NGTDM were independent predictors of adverse outcomes, with high hazard ratios. Two studies demonstrated a high ability to identify primary refractory disease (AUC > 0.95). Integrating multiple lesions increased predictive accuracy compared with the analysis of a single target lesion. Substantial variability persisted in feature segmentation, extraction, and standardization.

Baseline ¹⁸F-FDG PET/CT-based radiomics provides relevant complementary prognostic information in HL, particularly for predicting survival and treatment refractoriness. However, methodological heterogeneity, small sample sizes, and the lack of external validation currently limit its clinical implementation. For radiomics to be realistically incorporated into clinical practice, further studies are needed that are externally validated in large, multicenter cohorts, methodologically reproducible, and widely available for use and validation.