Abstract P078

Baseline and Delta PET/CT Radiomics for Prognostic Stratification in Paediatric Hodgkin Lymphoma

Background: Reliable prognostic biomarkers are needed to improve risk stratification and treatment adaptation in paediatric Hodgkin lymphoma (HL). While conventional PET metrics such as SUVmax and total metabolic tumor volume (TMTV) are clinically established, the added value of longitudinal radiomic heterogeneity features remains incompletely understood. This study investigated the prognostic significance of baseline and delta PET/CT radiomic features for progression-free survival (PFS) in paediatric HL.

Methods: A retrospective cohort of 61 paediatric HL patients undergoing FDG PET/CT imaging was analyzed. Clinical, PET, CT, and radiomic features were extracted from baseline and paired longitudinal scans. Baseline metrics included SUVmax, TMTV, total lesion glycolysis (TLG), DMaxBulk, entropy-based texture features, and neighborhood gray-tone difference matrix (NGTDM) features. Longitudinal delta features were computed between sequential scans. Cox proportional hazards analysis, Kaplan–Meier survival analysis (log-rank test), ROC analysis, and subgroup analyses for early- and advanced-stage disease were performed. Optimal prognostic cutoffs were determined using the Youden index.

Results: On univariate analysis (UVA), baseline TMTV and DMaxBulk demonstrated significant associations with PFS (p=0.016 and p=0.001, respectively), whereas baseline SUVmax was not predictive (p=0.757). CT NGTDM Complexity showed strong prognostic value (p=0.001), while baseline PET and CT entropy demonstrated borderline significance. ROC analysis with Youden-index optimization identified prognostic cutoffs for MTV, DMaxBulk, TLG, SUVmax, body surface area, and chest ratio, with corresponding Kaplan–Meier stratification analyses confirming significant survival separation for selected biomarkers. On univariate analysis, longitudinal analysis revealed that delta TMTV remained significantly associated with outcome (p=0.016), whereas changes in SUVmax were not predictive (p=0.631). Delta PET NGTDM Strength demonstrated significant prognostic value (p=0.001), supporting the importance of temporal heterogeneity assessment. Subgroup analyses showed that standard early response assessment retained prognostic significance in early-stage patients (p=0.025), while several radiomic biomarkers demonstrated stronger predictive performance in advanced-stage disease. Multivariable modeling suggests incremental prognostic value when combining radiomic biomarkers, clinical variables, and treatment-