Related Experiment Video
Updated: Jul 17, 2026
10:04
Analysis of 18FDG PET/CT Imaging as a Tool for Studying Mycobacterium tuberculosis Infection and Treatment in Non-human Primates
Published on: September 5, 2017
19.2K
Improving the predictive value of end-of-treatment PET/ CT in diffuse large B-cell lymphoma
Anne L Bes1, Gerben J C Zwezerijnen2, Martijn W Heymans3
1Department of Hematology, Amsterdam UMC, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands; 1Imaging and Biomarkers, Cancer Center Amsterdam, Amsterdam. a.l.bes@amsterdamumc.nl.
Haematologica
|January 8, 2026
Summary
Quantitative PET parameters improve diffuse large B-cell lymphoma treatment response prediction. New models enhance positive predictive value (PPV) and reduce false positives, aiding treatment decisions for PET/CT scans.
Area of Science:
- Oncology
- Nuclear Medicine
- Radiomics
Background:
- The 5-point Deauville score (DS) assesses end-of-treatment (EOT) response in diffuse large B-cell lymphoma (DLBCL) using PET/CT.
- The current DS has a suboptimal positive predictive value (PPV) of 60%, leading to potential overtreatment.
- Improving the accuracy of EOT response assessment is crucial for personalized treatment strategies in DLBCL.
Purpose of the Study:
- To evaluate if quantitative PET parameters and clinical data can enhance treatment failure prediction in EOT PET-positive DLBCL patients.
- To develop and validate predictive models for 2-year progression-free survival in DLBCL patients with DS 4-5 scans.
Main Methods:
- Analysis of baseline and EOT PET/CT scans from 138 DLBCL patients with DS 4-5.
- Segmentation of lesions using a semi-automated adaptive method (SUV4.0 or MV3).
- Extraction of PET parameters: total metabolic tumor volume (TMTV), number of lesions (NOL), tumorSUV/liverSUV-ratio (TLR), DmaxBulk.
- Development of two Cox regression models: Model 1 (clinical data + EOT PET), Model 2 (baseline, EOT, and delta PET values).
- Internal bootstrapping for model validation and performance evaluation (sensitivity, specificity, PPV, NPV).
Main Results:
- Model 1 included NOL and EOT TLRpeakmean (c-index=0.747); Model 2 included NOL, EOT TLRpeakmean, and baseline SUVmean (c-index=0.762).
- Both models significantly improved PPV to over 85% without compromising negative predictive value (NPV).
- False positives decreased from 39% (54 patients) with DS to 7% (9 patients) with Model 1 and 4% (6 patients) with Model 2.
- Inclusion of baseline features did not substantially alter model performance.
Conclusions:
- Quantitative PET parameters combined with clinical data or baseline PET features can significantly improve the prediction of treatment failure in EOT PET-positive DLBCL patients.
- These enhanced models reduce false positives, potentially avoiding unnecessary treatments.
- The developed models offer a valuable tool for more accurate response-adapted treatment decisions in DLBCL management.
