Jakoba Johanna Eertink

Purpose: SUV4.0-based thresholding is widely used for baseline [¹⁸F]FDG PET-based metabolic tumor volume (MTV) assessment in diffuse large B-cell lymphoma (DLBCL), but its suitability at interim and end-of-treatment (EoT) PET, when residual uptake is heterogeneous and tumor-to-background contrast is lower, is uncertain. We aimed to define a lesion-adaptive decision rule approach for selecting the optimal segmentation method based on lesion-level features and treatment phase and, exploratorily, to compare its performance with ML-based selection models. Methods: A total of 598 lesions from 33 DLBCL patients (HOVON-84 trial) were segmented at baseline, interim, and EoT [¹⁸F]FDG PET/CT using six semi-automated methods: SUV2.5, SUV4.0, 41%max, A50peak, MV2, and MV3. Segmentation quality was independently rated for each lesion by two observers (scale 1–5; 3 = preferred), with adjudication by a third reviewer. The influence of lesional SUVpeak, tumor-to-background ratio (TBRpeak), background uptake (SUVbg), treatment phase, and location on segmentation quality was assessed. Over six million rule-based combinations of key features were evaluated to derive a lesion-adaptive decision rule for preferred method selection. Exploratorily, ML classifiers were trained and compared with the decision-rule strategy. Results: Segmentation quality varied across lesions and methods. SUVpeak, TBRpeak, and SUVbg were key predictors of method performance. The final lesion-adaptive rule, applying SUV4.0 if SUVpeak > 8, MV3 if SUVbg > 0.8, and otherwise MV2, achieved a lesion-wise accuracy of 0.82 for preferred method selection, matching the best-performing ML models. Versus SUV4.0 alone (benchmark), the Decision Rule improved lesion-level MTV agreement with the reference (ρ = 0.85 vs. 0.82 vs. best ML ρ = 0.81) and reduced the proportion of lesions with > 10% MTV deviation (46.2% vs. 63.5%; best ML 50.2%). Total-MTV agreements with the reference were uniformly high across all strategies (all ρ ≥ 0.94), with modest gains for the decision rule at interim and EoT PET. Conclusion: A straightforward decision-rule approach using SUVpeak and SUVbg successfully selects the preferred method for individual DLBCL lesions across treatment phases and matches ML performance with greater simplicity and clinical applicability. Although supervision remains necessary, this approach helps address the current gap in segmentation methodology for interim and EoT PET, where SUV4.0 may not always be appropriate. Graphical

Purpose: Diffuse large B-cell lymphoma (DLBCL) patients often experience relapse or progression after treatment, emphasizing the need for better prognostic markers. This study investigated baseline FDG uptake and volume of healthy tissues and tumours, to assess their association with time-to-progression (TTP) in DLBCL. Methods: This study included 259 newly diagnosed DLBCL patients. Outcome was two-year TTP after treatment initiation. Automatic segmentation of tissues was performed on the low dose CT scans. Tumour lesions were outlined with a semi-automated segmentation method (standardised uptake value ≥ 4) and subtracted from the tissues. Mean standardised uptake value (SUVmean), tissue-to-blood SUVmean ratio, volume, and total lesion glycolysis (TLG) were determined for the spleen, liver, kidneys, lungs, and brain. Only SUVmean and SUVmean ratio were assessed for fat, bone, and skeletal muscle. Intercorrelations among all tissues and correlations with tumour TLG were calculated. Volume and uptake measures were compared between patients with and without progression using the Wilcoxon signed-rank test. Any measure significantly associated with two-year TTP was added to an existing logistic regression model based on baseline tumour characteristics to assess its added prognostic value. Results: Patients with progression showed a significantly higher spleen-to-blood SUVmean ratio, lower SUVmean and TLG in the brain, lower SUVmean in skeletal muscle, and a larger liver volume. Liver volume and spleen-to-blood SUVmean ratio did not significantly improve the prediction of two-year TTP compared to the original model only based on tumour characteristics. Conclusion: Healthy tissue PET/CT measures were significantly associated with two-year TTP in DLBCL patients. However, these measures did not improve the prediction of two-year TTP compared to models using only tumour characteristics. Trial registration number and date of registration: HOVON-84: EudraCT: 2006-005, 174 − 42, retrospectively registered 01-08-2008.

Immune-based treatment approaches are successfully used for the treatment of patients with cancer. While such therapies can be highly effective, many patients fail to benefit. To provide optimal therapy choices and to predict treatment responses, reliable biomarkers for the assessment of immune features in patients with cancer are of significant importance. Biomarkers (BM) that enable a comprehensive and repeatable assessment of the tumor microenvironment (TME), the lymphoid system, and the dynamics induced by drug treatment can fill this gap. Medical imaging, notably positron emission tomography (PET) and magnetic resonance imaging (MRI), providing whole-body imaging BMs, might deliver such BMs. However, those imaging BMs must be well characterized as being ‘fit for purpose’ for the intended use. This review provides an overview of the key steps involved in the development of ‘fit-for-purpose’ imaging BMs applicable in drug development, with a specific focus on pharmacodynamic biomarkers for assessing the TME and its modulation by immunotherapy. The importance of the qualification of imaging BMs according to their context of use (COU) as defined by the Food and Drug Administration (FDA) and National Institutes of Health Biomarkers, EndpointS, and other Tools (BEST) glossary is highlighted. We elaborate on how an imaging BM qualification for a specific COU can be achieved.

Investigating prognostic factors in patients with relapsed or primary refractory classical Hodgkin lymphoma (R/R cHL) is essential to optimize risk-adapted treatment strategies. We built a prognostic model using baseline quantitative 18F-fluorodeoxyglucose positron emission tomography (PET) radiomics features and clinical characteristics to predict the progression-free survival (PFS) among patients with R/R cHL treated with salvage chemotherapy followed by autologous stem cell transplantation. Metabolic tumor volume and several novel radiomics dissemination features, representing interlesional differences in distance, volume, and standard uptake value, were extracted from the baseline PET. Machine learning using backward selection and logistic regression were applied to develop and train the model on a total of 113 patients from 2 clinical trials. The model was validated on an independent external cohort of 69 patients. In addition, we validated 4 different PET segmentation methods to calculate radiomics features. We identified a subset of patients at high risk for progression with significant inferior 3-year PFS outcomes of 38.1% vs 88.4% for patients in the low-risk group in the training cohort (P < .001) and 38.5% vs 75.0% in the validation cohort (P = .015), respectively. The overall survival was also significantly better in the low-risk group (P = .022 and P < .001). We provide a formula to calculate a risk score for individual patients based on the model. In conclusion, we developed a prognostic model for PFS combining radiomics and clinical features in a large cohort of patients with R/R cHL. This model calculates a PET-based risk profile and can be applied to develop risk-stratified treatment strategies for patients with R/R cHL. These trials were registered at www.clinicaltrials.gov as #NCT02280993, #NCT00255723, and #NCT01508312.

The objective of this study was to externally validate the clinicalPET model developed in the HOVON-84 trial and to compare the model performance of our clinicalPET model to the international prognostic index (IPI). In total, 1195 Diffuse large B-cell lymphoma patients were included. 887 patients from 6 studies were used as external validation datasets. Primary outcomes were 2-year progression free survival (PFS) and 2-year time to progression (TTP). Metabolic tumor volume (MTV), the maximum distance between the largest lesion and another lesion (Dmaxbulk) and the peak standardized uptake value (SUVpeak) were extracted. The predictive value of the IPI and the HOVON-84 clinicalPET model (MTV, Dmaxbulk, SUVpeak, performance status and age) were tested. Model performance was assessed using the area under the curve (AUC), and diagnostic performance with the positive predictive value (PPV). Using 2-year PFS as outcome, the IPI yielded an AUC of 0.62 (range:0.51-0.65). The clinicalPET model yielded a significantly higher AUC of 0.71 (range:0.59-0.75, p<0.001). Comparable results were found using 2-year TTP. High-risk IPI patients had a 2-year PFS of 61.4%, versus 51.9% for the high-risk clinicalPET patients, with an increase in PPV from 35.5% to 49.1%, respectively. 66.4% of high-risk IPI patients were free from progression or relapse versus 55.5% for high-risk clinicalPET patients, with an increased PPV from 33.7% to 44.6%, respectively. The clinicalPET model that was developed in the HOVON-84 dataset remained predictive of outcome in 6 independent first-line DLBCL studies, and had higher model performance than the currently used IPI in all studies.

We investigated whether the outcome prediction of patients with aggressive B-cell lymphoma can be improved by combining clinical, molecular genotype, and radiomics features. MYC, BCL2, and BCL6 rearrangements were assessed using fluorescence in situ hybridization. Seventeen radiomics features were extracted from the baseline positron emission tomography–computed tomography of 323 patients, which included maximum standardized uptake value (SUVmax), SUVpeak, SUVmean, metabolic tumor volume (MTV), total lesion glycolysis, and 12 dissemination features pertaining to distance, differences in uptake and volume between lesions, respectively. Logistic regression with backward feature selection was used to predict progression after 2 years. The predictive value of (1) International Prognostic Index (IPI); (2) IPI plus MYC; (3) IPI, MYC, and MTV; (4) radiomics; and (5) MYC plus radiomics models were tested using the cross-validated area under the curve (CV-AUC) and positive predictive values (PPVs). IPI yielded a CV-AUC of 0.65 ± 0.07 with a PPV of 29.6%. The IPI plus MYC model yielded a CV-AUC of 0.68 ± 0.08. IPI, MYC, and MTV yielded a CV-AUC of 0.74 ± 0.08. The highest model performance of the radiomics model was observed for MTV combined with the maximum distance between the largest lesion and another lesion, the maximum difference in SUVpeak between 2 lesions, and the sum of distances between all lesions, yielding an improved CV-AUC of 0.77 ± 0.07. The same radiomics features were retained when adding MYC (CV-AUC, 0.77 ± 0.07). PPV was highest for the MYC plus radiomics model (50.0%) and increased by 20% compared with the IPI (29.6%). Adding radiomics features improved model performance and PPV and can, therefore, aid in identifying poor prognosis patients.

Introduction: Although visual and quantitative assessments of [18F]FDG PET/CT studies typically rely on liver uptake value as a reference or normalisation factor, consensus or consistency in measuring [18F]FDG uptake is lacking. Therefore, we evaluate the variation of several liver standardised uptake value (SUV) measurements in lymphoma [18F]FDG PET/CT studies using different uptake metrics. Methods: PET/CT scans from 34 lymphoma patients were used to calculate SUVmax^liver, SUVpeak^liver and SUVmean^liver as a function of (1) volume-of-interest (VOI) size, (2) location, (3) imaging time point and (4) as a function of total metabolic tumour volume (MTV). The impact of reconstruction protocol on liver uptake is studied on 15 baseline lymphoma patient scans. The effect of noise on liver SUV was assessed using full and 25% count images of 15 lymphoma scans. Results: Generally, SUVmax^liver and SUVpeak^liver were 38% and 16% higher compared to SUVmean^liver. SUVmax^liver and SUVpeak^liver increased up to 31% and 15% with VOI size while SUVmean^liver remained unchanged with the lowest variability for the largest VOI size. Liver uptake metrics were not affected by VOI location. Compared to baseline, liver uptake metrics were 15–18% and 9–18% higher at interim and EoT PET, respectively. SUV^liver decreased with larger total MTVs. SUVmax^liver and SUVpeak^liver were affected by reconstruction protocol up to 62%. SUVmax and SUVpeak moved 22% and 11% upward between full and 25% count images. Conclusion: SUVmean^liver was most robust against VOI size, location, reconstruction protocol and image noise level, and is thus the most reproducible metric for liver uptake. The commonly recommended 3 cm diameter spherical VOI-based SUVmean^liver values were only slightly more variable than those seen with larger VOI sizes and are sufficient for SUVmean^liver measurements in future studies. Trial registration: EudraCT: 2006–005,174-42, 01–08-2008.

Aim: Clinical prediction models need to be validated. In this study, we used simulation data to compare various internal and external validation approaches to validate models. Methods: Data of 500 patients were simulated using distributions of metabolic tumor volume, standardized uptake value, the maximal distance between the largest lesion and another lesion, WHO performance status and age of 296 diffuse large B cell lymphoma patients. These data were used to predict progression after 2 years based on an existing logistic regression model. Using the simulated data, we applied cross-validation, bootstrapping and holdout (n = 100). We simulated new external datasets (n = 100, n = 200, n = 500) and simulated stage-specific external datasets (1), varied the cut-off for high-risk patients (2) and the false positive and false negative rates (3) and simulated a dataset with EARL2 characteristics (4). All internal and external simulations were repeated 100 times. Model performance was expressed as the cross-validated area under the curve (CV-AUC ± SD) and calibration slope. Results: The cross-validation (0.71 ± 0.06) and holdout (0.70 ± 0.07) resulted in comparable model performances, but the model had a higher uncertainty using a holdout set. Bootstrapping resulted in a CV-AUC of 0.67 ± 0.02. The calibration slope was comparable for these internal validation approaches. Increasing the size of the test set resulted in more precise CV-AUC estimates and smaller SD for the calibration slope. For test datasets with different stages, the CV-AUC increased as Ann Arbor stages increased. As expected, changing the cut-off for high risk and false positive- and negative rates influenced the model performance, which is clearly shown by the low calibration slope. The EARL2 dataset resulted in similar model performance and precision, but calibration slope indicated overfitting. Conclusion: In case of small datasets, it is not advisable to use a holdout or a very small external dataset with similar characteristics. A single small testing dataset suffers from a large uncertainty. Therefore, repeated CV using the full training dataset is preferred instead. Our simulations also demonstrated that it is important to consider the impact of differences in patient population between training and test data, which may ask for adjustment or stratification of relevant variables.

Purpose: Biomarkers that can accurately predict outcome in DLBCL patients are urgently needed. Radiomics features extracted from baseline [¹⁸F]-FDG PET/CT scans have shown promising results. This study aims to investigate which lesion- and feature-selection approaches/methods resulted in the best prediction of progression after 2 years. Methods: A total of 296 patients were included. 485 radiomics features (n = 5 conventional PET, n = 22 morphology, n = 50 intensity, n = 408 texture) were extracted for all individual lesions and at patient level, where all lesions were aggregated into one VOI. 18 features quantifying dissemination were extracted at patient level. Several lesion selection approaches were tested (largest or hottest lesion, patient level [all with/without dissemination], maximum or median of all lesions) and compared to the predictive value of our previously published model. Several data reduction methods were applied (principal component analysis, recursive feature elimination (RFE), factor analysis, and univariate selection). The predictive value of all models was tested using a fivefold cross-validation approach with 50 repeats with and without oversampling, yielding the mean cross-validated AUC (CV-AUC). Additionally, the relative importance of individual radiomics features was determined. Results: Models with conventional PET and dissemination features showed the highest predictive value (CV-AUC: 0.72–0.75). Dissemination features had the highest relative importance in these models. No lesion selection approach showed significantly higher predictive value compared to our previous model. Oversampling combined with RFE resulted in highest CV-AUCs. Conclusion: Regardless of the applied lesion selection or feature selection approach and feature reduction methods, patient level conventional PET features and dissemination features have the highest predictive value. Trial registration number and date: EudraCT: 2006–005174-42, 01–08-2008.

Blood‐based biomarkers are gaining interest for response evaluation in classical Hodgkin lymphoma (cHL). However, it is unknown how blood‐based biomarkers relate to quantitative ¹⁸F‐FDG‐PET features. We correlated extracellular vesicle‐associated miRNAs (EV‐miRNA), serum TARC, and complete blood count (CBC) with PET features (e.g., metabolic tumor volume [MTV], dissemination and intensity features) in 30 cHL patients at baseline. EV‐miR127‐3p, EV‐miR24‐3p, sTARC, and several CBC parameters showed weak to strong correlations with MTV and dissemination features, but not with intensity features. Two other EV‐miRNAs only showed weak correlations with PET features. Therefore, blood‐based biomarkers may be complementary to PET features, which warrants further exploration of combining these biomarkers in prognostic models.