A PET-radiomics diagnostic signature to differentiate Hodgkin lymphoma from aggressive non-Hodgkin lymphomas.

OBJECTIVES

Noninvasive differentiation between Hodgkin lymphoma and aggressive non-Hodgkin lymphoma remains clinically important yet challenging. We developed and internally validated a PET-based radiomic model using routine baseline 18 F-FDG images to assist histological discrimination when tissue diagnosis is difficult.

METHODS

In this retrospective single-center study, baseline 18 F-FDG PET scans and clinical records from 190 histologically confirmed lymphoma patients were analyzed. Tumor volumes were segmented and high throughput radiomic features were extracted. Feature selection combined univariate filtering with a sparsity-promoting algorithm, and a multivariable logistic regression produced the final radiomics signature.

RESULTS

A compact radiomic signature that captures intratumoral heterogeneity separated Hodgkin lymphoma from aggressive non-Hodgkin lymphoma. On an independent validation cohort, the model demonstrated good discriminatory ability (AUC=0.71), and performance metrics indicate potential clinical utility as a noninvasive adjunct to standard assessment.

CONCLUSIONS

Radiomics applied to routine baseline 18 F-FDG PET can provide a practical, image-based biomarker to support lymphoma subtyping when biopsy is limited. Prospective, multi-center validation is recommended before clinical deployment.