Prognostic Value of Volumetric-Metabolic Parameters on 18F-FDG Positron Emission Tomography in Diffuse Large B-Cell Lymphoma: a Comparative Analysis of Segmentation Methods

Introduction : Diffuse large B-cell lymphoma (DLBCL) is the most common form of aggressive non-Hodgkin͛s lymphoma, accounting for up to 40 % of all cases. Despite significant advances in chemotherapy and immunotherapy, long-term treatment outcomes are not always satisfactory. In recent years, quantitative metrics derived from 18 F-fluorodeoxyglucose positron emission tomography ( 18 F-FDG PET/CT), such as metabolic tumor volume (MTV), total glycolytic tumor volume (TLG), and others, have received increasing attention as potential biomarkers for risk stratification and assessment of treatment response. Purpose : To evaluate the prognostic value of 18 F-FDG PET/CT volumetric-metabolic parameters in patients with diffuse large B-cell lymphoma (DLBCL) and to compare the effectiveness of different tumor burden segmentation methods. Materials and methods : Data from 150 DLBCL patients treated with R-CHOP regimen between 2018 and 2024 were included. All patients underwent 18 F-FDG PET/CT prior to therapy initiation. Metabolic tumor volume (MTV) and total lesion glycolysis (TLG) were calculated using two segmentation thresholds: a fixed threshold (SUV > 4) and a relative threshold (41 % of SUV max ). Receiver operating characteristic (ROC) analysis, machine learning techniques (yGBoost), and progression-free survival (PFS) assessment were used to evaluate prognostic significance. Results : All investigated parameters (MTV, TLG, and their normalized counterparts) demonstrated comparable prognostic accuracy (AUC 0.766–0.790). The body weight-normalized MTV using the SUV > 4 threshold (sMTV) showed the highest prognostic significance, enabling patient stratification into low-risk (< 3 mL/kg) and high-risk (≥​​3 mL/kg) groups. The 5-year PFS rates were 83.3 % and 39.4 %, respectively (p < 0.001). Discussion : Traditional methods such as ROC analysis and the DeLong test do not account for interactions, making it difficult to select the best metric. In this study, the yGBoost algorithm identified the weight-normalized volume, calculated using the SUV > 4 threshold (sMTV SUV > 4), as the most significant predictor. Conclusion : Body weight-normalized metabolic tumor volume is the most informative prognostic marker in DLBCL patients, highlighting its potential clinical value for personalized treatment approaches.