Pet-radiomics in lymphoma and multiple myeloma: update of current literature

Purpose To provide a comprehensive overview of the current literature on the applications of PET-based radiomics in patients affected by multiple myeloma (MM) and FDG-avid lymphomas. Methods Relevant studies on the topic were selected by searching Pubmed/Medline. Retrospective or prospective cohort studies focusing on the clinical applications of PET-radiomics in lymphomas and MM were retrieved, analyzed, and discussed. Result A total of 17 papers were ultimately selected, with 9 focusing on non-Hodgkin lymphomas, 6 on Hodgkin lymphomas, and 5 dealing with MM. Machine learning-derived models incorporating first-, second-, and third-order radiomic features extracted from baseline PET/CT scans demonstrated promising results in predicting outcomes, specifically the 2-year event-free survival (EFS) in lymphomas. Furthermore, models based on PET-radiomic features were effective in distinguishing between MM and bone metastases, as well as in assessing minimal residual disease, outperforming visual analysis. Conclusion Preliminary results suggest that PET-radiomic features, which reflect the biological heterogeneity and spatial distribution of lesions, may play a prognostic role in both lymphomas and MM. Nevertheless, before implementing these findings in clinical practice, it is imperative to standardize the methodological approaches and validate them in large prospective trials.