Sensitivity of an AI method for [18F]FDG PET/CT outcome prediction of diffuse large B-cell lymphoma patients to image reconstruction protocols.

BACKGROUND: Convolutional neural networks (CNNs), applied to baseline [18F]-FDG PET/CT maximum intensity projections (MIPs), show potential for treatment outcome prediction in diffuse large B-cell lymphoma (DLBCL). The aim of this study is to investigate the robustness of CNN predictions to different image reconstruction protocols. Baseline [18F]FDG PET/CT scans were collected from 20 DLBCL patients. EARL1, EARL2 and high-resolution (HR) protocols were applied per scan, generating three images with different image qualities. Image-based transformation was applied by blurring EARL2 and HR images to generate EARL1 compliant images using a Gaussian filter of 5 and 7 mm, respectively. MIPs were generated for each of the reconstructions, before and after image transformation. An in-house developed CNN predicted the probability of tumor progression within 2 years for each MIP. The difference in probabilities per patient was then calculated between both EARL2 and HR with respect to EARL1 (delta probabilities or ΔP). We compared these to the probabilities obtained after aligning the data with ComBat using the difference in median and interquartile range (IQR). RESULTS: CNN probabilities were found to be sensitive to different reconstruction protocols (EARL2 ΔP: median = 0.09, interquartile range (IQR) = [0.06, 0.10] and HR ΔP: median = 0.1, IQR = [0.08, 0.16]). Moreover, higher resolution images (EARL2 and HR) led to higher probability values. After image-based and ComBat transformation, an improved agreement of CNN probabilities among reconstructions was found for all patients. This agreement was slightly better after image-based transformation (transformed EARL2 ΔP: median = 0.022, IQR = [0.01, 0.02] and transformed HR ΔP: median = 0.029, IQR = [0.01, 0.03]). CONCLUSION: Our CNN-based outcome predictions are affected by the applied reconstruction protocols, yet in a predictable manner. Image-based harmonization is a suitable approach to harmonize CNN predictions across image reconstruction protocols.

An artificial intelligence method using FDG PET to predict treatment outcome in diffuse large B cell lymphoma patients.

Convolutional neural networks (CNNs) may improve response prediction in diffuse large B-cell lymphoma (DLBCL). The aim of this study was to investigate the feasibility of a CNN using maximum intensity projection (MIP) images from 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography (PET) baseline scans to predict the probability of time-to-progression (TTP) within 2 years and compare it with the International Prognostic Index (IPI), i.e. a clinically used score. 296 DLBCL 18F-FDG PET/CT baseline scans collected from a prospective clinical trial (HOVON-84) were analysed. Cross-validation was performed using coronal and sagittal MIPs. An external dataset (340 DLBCL patients) was used to validate the model. Association between the probabilities, metabolic tumour volume and Dmaxbulk was assessed. Probabilities for PET scans with synthetically removed tumors were also assessed. The CNN provided a 2-year TTP prediction with an area under the curve (AUC) of 0.74, outperforming the IPI-based model (AUC = 0.68). Furthermore, high probabilities (> 0.6) of the original MIPs were considerably decreased after removing the tumours (< 0.4, generally). These findings suggest that MIP-based CNNs are able to predict treatment outcome in DLBCL.

Baseline radiomics features and MYC rearrangement status predict progression in aggressive B-cell lymphoma.

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.

Reproducibility of [18F]FDG PET/CT liver SUV as reference or normalisation factor.

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 SUVmaxliver, SUVpeakliver and SUVmeanliver 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, SUVmaxliver and SUVpeakliver were 38% and 16% higher compared to SUVmeanliver. SUVmaxliver and SUVpeakliver increased up to 31% and 15% with VOI size while SUVmeanliver 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. SUVliver decreased with larger total MTVs. SUVmaxliver and SUVpeakliver were affected by reconstruction protocol up to 62%. SUVmax and SUVpeak moved 22% and 11% upward between full and 25% count images. CONCLUSION: SUVmeanliver 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 SUVmeanliver values were only slightly more variable than those seen with larger VOI sizes and are sufficient for SUVmeanliver measurements in future studies. TRIAL REGISTRATION: EudraCT: 2006-005,174-42, 01-08-2008.

Comparing lesion and feature selections to predict progression in newly diagnosed DLBCL patients with FDG PET/CT radiomics features.

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.

Combatting the effect of image reconstruction settings on lymphoma [18F]FDG PET metabolic tumor volume assessment using various segmentation methods.

BACKGROUND: [18F]FDG PET-based metabolic tumor volume (MTV) is a promising prognostic marker for lymphoma patients. The aim of this study is to assess the sensitivity of several MTV segmentation methods to variations in image reconstruction methods and the ability of ComBat to improve MTV reproducibility. METHODS: Fifty-six lesions were segmented from baseline [18F]FDG PET scans of 19 lymphoma patients. For each scan, EARL1 and EARL2 standards and locally clinically preferred reconstruction protocols were applied. Lesions were delineated using 9 semiautomatic segmentation methods: fixed threshold based on standardized uptake value (SUV), (SUV = 4, SUV = 2.5), relative threshold (41% of SUVmax [41M], 50% of SUVpeak [A50P]), majority vote-based methods that select voxels detected by at least 2 (MV2) and 3 (MV3) out of the latter 4 methods, Nestle thresholding, and methods that identify the optimal method based on SUVmax (L2A, L2B). MTVs from EARL2 and locally clinically preferred reconstructions were compared to those from EARL1. Finally, different versions of ComBat were explored to harmonize the data. RESULTS: MTVs from the SUV4.0 method were least sensitive to the use of different reconstructions (MTV ratio: median = 1.01, interquartile range = [0.96-1.10]). After ComBat harmonization, an improved agreement of MTVs among different reconstructions was found for most segmentation methods. The regular implementation of ComBat ('Regular ComBat') using non-transformed distributions resulted in less accurate and precise MTV alignments than a version using log-transformed datasets ('Log-transformed ComBat'). CONCLUSION: MTV depends on both segmentation method and reconstruction methods. ComBat reduces reconstruction dependent MTV variability, especially when log-transformation is used to account for the non-normal distribution of MTVs.

Proposed New Dynamic Prognostic Index for Diffuse Large B-Cell Lymphoma: International Metabolic Prognostic Index.

PURPOSE: Baseline metabolic tumor volume (MTV) is a promising biomarker in diffuse large B-cell lymphoma (DLBCL). Our aims were to determine the best statistical relationship between MTV and survival and to compare MTV with the International Prognostic Index (IPI) and its individual components to derive the best prognostic model. METHODS: PET scans and clinical data were included from five published studies in newly diagnosed diffuse large B-cell lymphoma. Transformations of MTV were compared with the primary end points of 3-year progression-free survival (PFS) and overall survival (OS) to derive the best relationship for further analyses. MTV was compared with IPI categories and individual components to derive the best model. Patients were grouped into three groups for survival analysis using Kaplan-Meier analysis; 10% at highest risk, 30% intermediate risk, and 60% lowest risk, corresponding with expected clinical outcome. Validation of the best model was performed using four studies as a test set and the fifth study for validation and repeated five times. RESULTS: The best relationship for MTV and survival was a linear spline model with one knot located at the median MTV value of 307.9 cm3. MTV was a better predictor than IPI for PFS and OS. The best model combined MTV with age as continuous variables and individual stage as I-IV. The MTV-age-stage model performed better than IPI and was also better at defining a high-risk group (3-year PFS 46.3% v 58.0% and 3-year OS 51.5% v 66.4% for the new model and IPI, respectively). A regression formula was derived to estimate individual patient survival probabilities. CONCLUSION: A new prognostic index is proposed using MTV, age, and stage, which outperforms IPI and enables individualized estimates of patient outcome.

Quantitative Radiomics Features in Diffuse Large B-Cell Lymphoma: Does Segmentation Method Matter?

Radiomics features may predict outcome in diffuse large B-cell lymphoma (DLBCL). Currently, multiple segmentation methods are used to calculate metabolic tumor volume (MTV). We assessed the influence of segmentation method on the discriminative power of radiomics features in DLBCL at the patient level and for the largest lesion. Methods: Fifty baseline 18F-FDG PET/CT scans of DLBCL patients with progression or relapse within 2 years after diagnosis were matched on uptake time and reconstruction method with 50 baseline PET/CT scans of DLBCL patients without progression. Scans were analyzed using 6 semiautomatic segmentation methods (SUV threshold of 4.0 [SUV4.0], SUV threshold of 2.5, 41% of SUVmax, 50% of SUVpeak, a majority vote segmenting voxels detected by ≥2 methods, and a majority vote segmenting voxels detected by ≥3 methods). On the basis of these segmentations, 490 radiomics features were extracted at the patient level, and 486 features were extracted for the largest lesion. To quantify the agreement between features extracted from different segmentation methods, the intraclass correlation (ICC) agreement was calculated for each method compared with SUV4.0. The feature space was reduced by deleting features that had high Pearson correlations (≥0.7) with the previously established predictors MTV or SUVpeak Model performance was assessed using stratified repeated cross validation with 5 folds and 2,000 repeats, yielding the mean receiver-operating-characteristics curve integral for all segmentation methods using logistic regression with backward feature selection. Results: The percentage of features yielding an ICC of at least 0.75, compared with the SUV4.0 segmentation, was lowest for 50% of SUVpeak both at the patient level and for the largest lesion, with 77.3% and 66.7% of the features yielding an ICC of at least 0.75, respectively. Features did not correlate strongly with MTV, with at least 435 features at the patient level and 409 features for the largest lesion for all segmentation methods having a correlation coefficient of less than 0.7. Features correlated strongly with SUVpeak (at least 190 at patient level and 134 for the largest lesion were uncorrelated to SUVpeak, respectively). Receiver-operating-characteristics curve integrals ranged between 0.69 ± 0.11 and 0.84 ± 0.09 at the patient level and between 0.69 ± 0.11 and 0.73 ± 0.10 at the lesion level. Conclusion: Even though there are differences in the actual radiomics feature values derived and selected features among segmentation methods, there is no substantial difference in the discriminative power of radiomics features among segmentation methods.

18F-FDG PET baseline radiomics features improve the prediction of treatment outcome in diffuse large B-cell lymphoma.

PURPOSE: Accurate prognostic markers are urgently needed to identify diffuse large B-Cell lymphoma (DLBCL) patients at high risk of progression or relapse. Our purpose was to investigate the potential added value of baseline radiomics features to the international prognostic index (IPI) in predicting outcome after first-line treatment. METHODS: Three hundred seventeen newly diagnosed DLBCL patients were included. Lesions were delineated using a semi-automated segmentation method (standardized uptake value ≥ 4.0), and 490 radiomics features were extracted. We used logistic regression with backward feature selection to predict 2-year time to progression (TTP). The area under the curve (AUC) of the receiver operator characteristic curve was calculated to assess model performance. High-risk groups were defined based on prevalence of events; diagnostic performance was assessed using positive and negative predictive values. RESULTS: The IPI model yielded an AUC of 0.68. The optimal radiomics model comprised the natural logarithms of metabolic tumor volume (MTV) and of SUVpeak and the maximal distance between the largest lesion and any other lesion (Dmaxbulk, AUC 0.76). Combining radiomics and clinical features showed that a combination of tumor- (MTV, SUVpeak and Dmaxbulk) and patient-related parameters (WHO performance status and age > 60 years) performed best (AUC 0.79). Adding radiomics features to clinical predictors increased PPV with 15%, with more accurate selection of high-risk patients compared to the IPI model (progression at 2-year TTP, 44% vs 28%, respectively). CONCLUSION: Prediction models using baseline radiomics combined with currently used clinical predictors identify patients at risk of relapse at baseline and significantly improved model performance. TRIAL REGISTRATION NUMBER AND DATE: EudraCT: 2006-005,174-42, 01-08-2008.

18F-FDG PET Improves Baseline Clinical Predictors of Response in Diffuse Large B-Cell Lymphoma: The HOVON-84 Study.

We aimed to determine the added value of baseline metabolic tumor volume (MTV) and interim PET (I-PET) to the age-adjusted international prognostic index (aaIPI) to predict 2-y progression-free survival (PFS) in diffuse large B-cell lymphoma. Secondary objectives were to investigate optimal I-PET response criteria (using Deauville score [DS] or quantitative change in SUVmax [ΔSUVmax] between baseline and I-PET4 [observational I-PET scans after 4 cycles of rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone administered in 2-wk intervals with intensified rituximab in the first 4 cycles [R(R)-CHOP14]). Methods: I-PET4 scans in the HOVON-84 (Hemato-Oncologie voor Volwassenen Nederland [Haemato Oncology Foundation for Adults in the Netherlands]) randomized clinical trial (EudraCT 2006-005174-42) were centrally reviewed using DS (cutoff, 4-5). Additionally, ΔSUVmax (prespecified cutoff, 70%) and baseline MTV were measured. Multivariable hazard ratio (HR), positive predictive value (PPV), and negative predictive value (NPV) were obtained for 2-y PFS. Results: In total, 513 I-PET4 scans were reviewed according to DS, and ΔSUVmax and baseline MTV were available for 367 and 296 patients. The NPV of I-PET ranged between 82% and 86% for all PET response criteria. Univariate HR and PPV were better for ΔSUVmax (4.8% and 53%, respectively) than for DS (3.1% and 38%, respectively). aaIPI and ΔSUVmax independently predicted 2-y PFS (HR, 3.2 and 5.0, respectively); adding MTV brought about a slight improvement. Low or low-intermediate aaIPI combined with a ΔSUVmax of more than 70% (37% of patients) yielded an NPV of 93%, and the combination of high-intermediate or high aaIPI and a ΔSUVmax of 70% or less yielded a PPV of 65%. Conclusion: In this study on diffuse large B-cell lymphoma, I-PET after 4 cycles of R(R)-CHOP14 added predictive value to aaIPI for 2-y PFS, and both were independent response biomarkers in a multivariable Cox model. We externally validated that ΔSUVmax outperformed DS in 2-y PFS prediction.

Interobserver Agreement on Automated Metabolic Tumor Volume Measurements of Deauville Score 4 and 5 Lesions at Interim 18F-FDG PET in Diffuse Large B-Cell Lymphoma.

Metabolic tumor volume (MTV) on interim PET (I-PET) is a potential prognostic biomarker for diffuse large B-cell lymphoma (DLBCL). Implementation of MTV on I-PET requires a consensus on which semiautomated segmentation method delineates lesions most successfully with least user interaction. Methods used for baseline PET are not necessarily optimal for I-PET because of lower lesional SUVs at I-PET. Therefore, we aimed to evaluate which method provides the best delineation quality for Deauville score (DS) 4-5 DLBCL lesions on I-PET at the best interobserver agreement on delineation quality and, second, to assess the effect of lesional SUVmax on delineation quality and performance agreement. Methods: DS 4-5 lesions from 45 I-PET scans were delineated using 6 semiautomated methods: a fixed SUV threshold of 2.5 g/cm3, a fixed SUV threshold of 4.0 g/cm3, an adaptive threshold corrected for source-to-local background activity contrast at 50% of the SUVpeak, 41% of SUVmax per lesion, a majority vote including voxels detected by at least 2 methods, and a majority vote including voxels detected by at least 3 methods (MV3). Delineation quality per MTV was rated by 3 independent observers as acceptable or nonacceptable. For each method, observer scores on delineation quality, specific agreement, and MTV were assessed for all lesions and per category of lesional SUVmax (<5, 5-10, >10). Results: In 60 DS 4-5 lesions on I-PET, MV3 performed best, with acceptable delineation in 90% of lesions and a positive agreement of 93%. Delineation quality scores and agreement per method strongly depended on lesional SUV: the best delineation quality scores were obtained using MV3 in lesions with an SUVmax of less than 10 and using SUV4.0 in more 18F-FDG-avid lesions. Consequently, overall delineation quality and positive agreement improved by applying the most preferred method per SUV category instead of using MV3 as the single best method. The MV3- and SUV4.0-derived MTVs of lesions with an SUVmax of more than 10 were comparable after exclusion of visually failed MV3 contouring. For lesions with an SUVmax of less than 10, MTVs using different methods correlated poorly. Conclusion: On I-PET, MV3 performed best and provided the highest interobserver agreement regarding acceptable delineations of DS 4-5 DLBCL lesions. However, delineation-method preference strongly depended on lesional SUV. Therefore, we suggest exploration of an approach that identifies the optimal delineation method per lesion as a function of tumor 18F-FDG uptake characteristics, that is, SUVmax.