Skip to main content
Download PDF
Download ePub

Clinical scoring systems, molecular subtypes and baseline [18F]FDG PET/CT image analysis for prognosis of diffuse large B-cell lymphoma

Cancer Imaging volume 24, Article number: 168 (2024) Cite this article

Abstract

Diffuse large B-cell lymphoma (DLBCL) is a highly heterogeneous hematological malignancy resulting in a range of outcomes, and the early prediction of these outcomes has important implications for patient management. Clinical scoring systems provide the most commonly used prognostic evaluation criteria, and the value of genetic testing has also been confirmed by in-depth research on molecular typing. [18F]-fluorodeoxyglucose positron emission tomography / computed tomography ([18F]FDG PET/CT) is an invaluable tool for predicting DLBCL progression. Conventional baseline image-based parameters and machine learning models have been used in prognostic FDG PET/CT studies of DLBCL; however, numerous studies have shown that combinations of baseline clinical scoring systems, molecular subtypes, and parameters and models based on baseline FDG PET/CT image may provide better predictions of patient outcomes and aid clinical decision-making in patients with DLBCL.

Background

Diffuse large B-cell lymphoma (DLBCL) is the most common non-Hodgkin lymphoma and accounts for 30% of all lymphomas. Although 60%−70% of patients with newly diagnosed DLBCL can be cured using the traditional standard therapy combining rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP), up to 30%−40% of patients will have disease that is refractory to this treatment or will have a relapse after an initial response [1,2,3,4,5,6]. Patients with refractory disease have a poor prognosis after salvage chemotherapy, but their outcomes may be greatly improved following Pola-R-CHP, which was approved by Food and Drug Administration (FDA) in 2023, immunotherapy or targeted therapies [5,6,7]. It is therefore necessary to construct an accurate model for predicting patient outcomes to enable early risk stratification and optimal treatment decisions for patients with DLBCL [8].

Clinical predictive indicators have been widely used for assessing DLBCL prognosis. Additionally, baseline FDG PET/CT plays an increasingly important role in predicting DLBCL outcomes. Since 2014, the International Conference on Malignant Lymphoma imaging consensus guidelines has recognized the use of FDG PET/CT for evaluating glucose metabolism in lymphoma lesions [9,10,11,12]. Other FDG-PET-derived parameters, such as metabolic tumor volume (MTV), total lesion glycolysis (TLG), and maximal distance between two farthest lesions (Dmax) may also predict DLBCL outcomes [13].

Numerous studies have used novel indicators and risk factors to construct new models to predict DLBCL progression. This review focuses on the important clinical scoring systems, molecular subtypes, and FDG PET/CT parameters related to DLBCL prognosis, to guide the selection of treatment regimens after the prediction of DLBCL outcomes.

Clinical scoring systems

The International Prognostic Index (IPI) clinical scoring system has been widely used for risk stratification and to select rational therapeutic strategies in patients with DLBCL since 1993 [14, 15]. IPI versions have been updated to stratify patient prognosis along with changes in DLBCL treatment methods, including the revised IPI (R-IPI), National Comprehensive Cancer Network IPI (NCCN-IPI), central nervous system IPI (CNS-IPI), and age-adjusted IPI (aa-IPI) (Table 1).

Table 1 Summary of the clinical scoring systems mentioned in this study
Full size table

IPI and aa-IPI

The IPI stratifies DLBCL into four discrete risk categories (low, low-intermediate, high-intermediate and high) with five clinical characteristics: age, lactate dehydrogenase level, number of extra-nodal sites, Ann Arbor stage, and Eastern Cooperative Oncology Group (ECOG) performance status; however, its stratified prognostic ability has been greatly reduced by gradual changes in DLBCL treatment methods [15,16,17].

The aa-IPI was developed for patients aged ≤ 60 years, who have notably different outcomes from older patients, and thus this is the age limit for the most-intensive experimental treatments for non-Hodgkin lymphoma. The aa-IPI involves three adverse prognostic factors: disease stages III-IV, high lactate dehydrogenase level, and ECOG performance status ≥ 2 [18].

R-IPI

Since the late 1990s, rituximab (R) added to CHOP for DLBCL has significantly improved survival among all risk groups; however, the IPI discrimination ability has declined, especially among higher-risk patients. The R-IPI was therefore developed to risk-stratify DLBCL patients treated with R-CHOP [15, 16, 19] into three risk groups: low [0], intermediate [1, 2], and high [3,4,5, 16]. Sehn et al. used the R-IPI to identify three distinct prognostic groups with very good (94%), good (79%), or poor (55%) overall survival (OS) (P < 0.001) [14]. More precise grouping can help doctors to balance efficacy against excessive toxicity.

NCCN-IPI

Neither the IPI nor the R-IPI can identify risk groups with < 50% chance of survival (4-year OS: IPI 59%, R-IPI 55%) [14, 20]. Pooled data showed a 5-year OS of approximately 50% in the IPI high-risk group, and an enhanced NCCN-IPI was therefore constructed to identify the above subgroups [19]. Unlike the IPI system, the NCCN-IPI regarded bone marrow, CNS, liver/gastrointestinal tract, and lung lesions as risk factors [21]. The IPI, R-IPI, and NCCN-IPI gave 5-year OS estimates with accuracy rates of 54%-88%, 61%-93%, and 49%-92%, respectively. The NCCN-IPI may be the best-performing scoring system, with similar ability of the R-IPI for discerning subgroups with favorable long-term survival and better ability than the IPI for detecting a less-heterogeneous high-risk group [15, 20, 22].

CNS-IPI

CNS infiltration occurs in 2%-10% of DLBCLs patients, and the CNS-IPI is developed for cases involving CNS relapse. The prognostic score includes IPI risk factors and involvement of the kidney and/or adrenal glands [23]. The CNS-IPI-predicted CNS relapse rates were 0.0%, 0.8%, and 13.8% for patients with low, intermediate, and high risk, respectively [24]. The CNS-IPI has thus been proposed as a prognostic tool to improve prospective validation and guide therapy [24, 25].

Compared with CNS-IPI alone, the combination model of CNS-IPI, such as the model based on high CNS-IPI score and ABC/unclassified cell of origin (COO) or based on CNS-IPI and the model incorporating images and clinical variables, could identified the high-risk population with a higher 2-year CNS-relapse probability (15.2% or 17.1% vs 8.9%) [26, 27].

IBPS and R/R-IPI

Scoring systems other than IPI-based systems also exist. The Inflammation-Based Prognosis Score (IBPS) was constructed from the systemic immune inflammation index, prognostic nutrition index, and modified Glasgow prognostic score, and generated C-indices for OS in training and validation cohorts of 0.844 and 0.828, respectively [28]. The relapsed/refractory-IPI(R/R-IPI) was constructed for relapsed/refractory DLBCL patients, using only age and front-line time to progression, with good C-indices in discovery (0.67) and validation sets (0.64, 0.68). This study provided a robust method with readily available clinical details to identify patients that should be considered for immediate treatment with the complex and costly chimeric antigen receptor T-cell therapy [29, 30].

The addition of new clinical indicators has been proposed to improve the predictive ability of existing scoring systems, including low serum cholesterol, uric acid, and apolipoprotein A-I, absolute lymphocyte/monocyte ratio, red blood cell distribution width, platelet count, and beta-2 microglobulin level [31,32,33].

Molecular subtypes

Although the IPI is easy to apply in clinical practice, it does not fully account for disease heterogeneity [8]. Gene expression profiling has identified DLBCL subgroups (activated B-cell-like [ABC], germinal-center B-cell-like [GCB], and unclassified) according to the cells of origin. ABC DLBCL is characterized by B-cell-receptor dependence, constitutive nuclear factor-κB activation, and interferon regulatory factor4 (IRF4) /MUM1 multiple myeloma oncogene1 (MUM1) expression, while GCB DLBCL is characterized by CD10 expression and BCL2 rearrangements [8, 34, 35]. Compared with GCB, ABC DLBCL has higher risk of relapse and inferior outcome following R-CHOP [35, 36].

The appropriate treatments based on the subtype classification can improve prognosis. Different subclassifications requiring fluorescence in situ hybridization testing to identify MYC, BCL2, and BCL6 rearrangements have been introduced to identify patients with increased risk profiles [37, 38]. MYC translocation is a strong adverse prognostic factor related to inferior OS and progression-free survival (PFS) [35]. Li et al. recommended bromodomain and extra-terminal protein family inhibitor therapy, either alone or in combination with other drugs, to improve the prognosis of patients with MYC expression [39].

Previous studies designated the 5%–15% of DLBCL cases with MYC, BLC2, and/or BCL6 translocations as DHL (MYC/BLC2, MYC/BCL6) and THL (MYC/BLC2/BCL6), respectively [40,41,42,43]. However, the 2022 World Health Organization (WHO) and International Consensus Classification (ICC) recommendations re-categorized MYC/BCL6 as “DLBCL, not otherwise specified” and MYC/BCL2 and MYC/BLC2/BCL6 as “DLBCL/high-grade B-cell lymphoma-MYC/BCL2” [34, 40, 41]. Based on these new classifications, a multicenter, retrospective study including 220 patients with DLBCL revealed that MYC/BCL6 patients had superior and longer OS than patients with MYC/BCL2-rearrangements and THL, and treatment intensification was associated with next treatment time and OS in patients with MYC/BCL2 and THL but no improvement in MYC/BCL6 patients [41]. More clinical trials are needed to confirm the optimal classification for DLBCL prognosis.

As mentioned above, molecular subclassification might predict clinical outcomes of current therapeutic strategies, with specific phenotypes enabling the development of precision therapies [35, 36]. A study of 412 patients with DLBCL identified 14 metabolism-associated genes characteristic of the immunosuppressive microenvironment and associated with prognosis. The resulting metabolism-associated prognosis risk model may facilitate personalized treatment strategies and provide the basis for further studies of metabolism-associated genes and the immune microenvironment [42].

R-CHOP therapy may be extended to include personalized treatment with agents targeting genes for DLBCL. A randomized phase II trial reported that R-CHOP-X including targeted Bruton’s tyrosine kinase inhibitors (ibrutinib), histone deacetylase inhibitors, demethylating agents (decitabine), and lenalidomide based on mutated MCD, BN2, EZB, TP53, and N1, resulted in significantly higher 2-year PFS and OS rates than R-CHOP [43]. Wang's et al. study demonstrated that high cyclin D2 (CCND2) expression in ABC DLBCL was an independent prognostic indicator of PFS, potentially promoting further research on CCND2 inhibition and R-CHOP combination therapy [44]. Acylglycerol kinase inhibitors represent another possible approach to enhance the efficacy of venetoclax (a highly selective BCL-2 inhibitor) [45], while other targeted agents include the anti-CD79b antibody–drug conjugate polatuzumab vedotin, and anti-CD19 chimeric antigen receptor T-cell products [37].

FDG PET/CT parameters

FDG PET/CT is an essential screening tool for DLBCL because it can reflect differential glycolytic activity between lesions and healthy tissue [46, 47]. Baseline PET metrics have demonstrated prognostic value in DLBCL in many studies, including a phase III clinical trial of obinutuzumab plus CHOP chemotherapy (GOYA) [48].

The standardized uptake value (SUV), as a semiquantitative measure of FDG retention, including SUVmax, SUVmean, and SUVpeak, quantifies the ratio of radioactivity at a given image location and the whole-body injected radioactivity [13, 49]. A systematic review of 25 studies from 2011 to 2020 concluded that SUV in baseline FDG PET/CT could not predict PFS or OS in DLBCL patients [13, 48, 50, 51]. However, other FDG PET/CT indicators including MTV, TLG, and Dmax could have prognostic value (Fig. 1).

Fig. 1
figure 1

Three semiquantitative parameters in coronal (a) and axial coronal images(b, c) of FDG PET/CT from one patient: SUV, MTV and TLG. Dmax, the distance between two lesions that are furthest apart in sagittal images (d) of FDG PET/CT

Full size image

MTV and TLG

MTV is the volume of disease contoured at a specified SUV threshold, with some semiautomated methods: a fixed SUV threshold of 2.5/4.0 g/cm3, 41% of SUVmax per lesion, a majority vote including voxels detected by at least 2/3 methods (MV2/3) and so on [52,53,54], while TLG is the sum of the products of each lesion’s MTV and SUVmean [13, 55, 56]. Some studies have compared those methods. SUV2.5 and SUV41% were recommended by El-Galaly et al., while SUV4.0 and MV2 were recommended by Barrington et al. [52, 53]. MV3 performed best in Zwezerijnen et al.’s research, with acceptable delineation in 90% of lesions and a positive agreement of 93%. It is worth noting that, in their study, delineation quality scores and agreement per method strongly depended on lesional SUV, which means that an approach that identifies the optimal delineation method per lesion as a function of tumor [18F]FDG uptake characteristics is required [54]. In actual situations, although MV3 performs well in some cases, the SUV2.5 and SUV41% methods are more commonly used in clinical practice due to their simplicity, ease of standardization, and extensive research support [49,50,51, 57].

A randomized trial demonstrated that high total metabolic tumor volume (TMTV) was significantly associated with shorter PFS and OS [58], while in the other two studies, R-CHOP resulted in significantly worse outcomes in patients with TMTV > 220 cm3 than in those with TMTV < 220 cm3 [57, 59]. With chimeric antigen receptor (CAR) T-cell therapy has emerged as an option for relapsed/ refractory (R/R) DLBCL, the high baseline TMTV has been proven as a predictor of early progression in the form of unfavorable OS [60]. Kostakoglu et al. showed that baseline TMTV and TLG were independent predictors of 4-year PFS in DLBCL patients after first-line immunochemotherapy [48]. Using a model combining baseline TLG and MTV, Ceriani et al. confirmed significantly poorer outcomes for both DLBCL and primary mediastinal B-cell lymphoma in patients at high risk of progression (P < 0.001), with no treatment failure in the low-risk group [56]. Most studies, however, only demonstrated that TLG was associated with survival, rather than being an independent predictor of PFS and OS, and its predictive value for DLBCL requires more in-depth research in large and multicenter studies.

MTV can serve as a single prognostic indicator and also improve the predictive reliability of prognoses based on other indicators, such as stage, IPI scores, and ECOG performance status. In 2020, Mikhaeel et al. proposed a new dynamic prognostic index for DLBCL: International Metabolic Prognostic Index (IMPI) composed of MTV, age, and stage which represents a significant advance for implementing MTV in lymphoma research [61,62,63]. For the patients with R/R DLBCL treated with CAR T-cell, Winkelmann et al. found that only IMPI showed a significant trend for PFS stratification (P = 0.030), while both IPI and IMPI didn’t show a significant association with OS after CAR T-cell [62]. Zhao et al. found that patients with low MTV had better 2-year PFS and OS than those with high MTV, especially in the low-intermediate-risk NCCN-IPI subgroup [64]. The phase III GOYA study demonstrated that patients with high TMTV and IPI had higher risks of relapse or progression than those with low TMTV and IPI (5-year PFS: 49.0% vs 74.3%) [48]. As an IPI scoring indicator, ECOG performance status has been proven to be an independent indicator of PFS and OS [57, 65]. For example, in Thieblemont et al.’s study, ECOG > 2 had a relatively high HR for PFS and OS in all three test sets [59]. Based on a positive net reclassification index for 4-year PFS and OS, Vercellino et al. concluded that a combined TMTV/ECOG variable had a higher model performance than the IPI [57], and the integrated model based on PET and tumor genotyping had a negative predictive value of 100% for disease progression or recurrence in the low-risk group.

Dmax

Dissemination features, including the distance between two most distant lesions (Dmaxpatient) and the distance between the largest lesion and most distant lesions (Dmaxbulk) were first proposed as DCBCL prognostic factors by Cottereau et al. in 2019 and have since been widely used [66,67,68]. Dmax was considered to be a better prognostic predictor for DLBCL, reflecting the extent of tumor invasion [66,67,68,69], while Dmaxpatient and Dmaxbulk were negative prognostic factors for 4-year PFS (P < 0.001) and OS [67]. Eertink et al. also concluded that dissemination features had better predictive value than other PET parameters for 2-year progression of DLBCL [70].

Early identification of high-risk DLBCL patients who are unlikely to be cured by R-CHOP is an important step in testing alternative treatment approaches and requires a well-developed risk-scoring approach [71]. Many studies combined Dmax and MTV as complementary prognostic factors for predicting PFS and OS, reflecting tumor spread and tumor burden, respectively [66,67,68, 70]. In two studies involving different cohorts, a combined model based on MTV and Dmax identified significant differences in 4-year PFS and OS rates among the three groups. Specifically, this model could identify a group of patients with a poor prognosis (two risk factors) even after R-CHOP therapy, for whom clinicians might consider alternative treatment approaches [67, 71]. Standardized Dmax (SDmax) is Dmax normalized by body surface area. PFS differed significantly among three risk groups based on MTV (P = 0.031) and SDmax (P = 0.001) in high-risk (NCCN-IPI ≥ 4) and low-risk (NCCN-IPI < 4) groups [66]. Eertink et al. found that the area under the curve (AUC) for a clinical PET model based on MTV, Dmaxbulk, SUVpeak, performance status, and age (AUC = 0.71) was significantly larger (P < 0.001) than that for the IPI (AUC = 0.62) [72].

Machine learning

Developments in image processing and analysis technology have led to the increasing application of computational software, fixed algorithms, and neural network models to analyze PET images to predict DLBCL progression [56, 73,74,75,76]. The main research methods currently include texture analysis, radiomics, and deep learning [73]. (Fig. 2).

Fig. 2
figure 2

The workflow of machine learning, which includes texture analysis, radiomics, and deep learning, consists of four steps. Firstly, manually or automatically delineate the lesions to obtain the region of interest. Secondly, translate images into radiomic features. Thirdly, select features associated with prognosis for model construction. Finally, validate the predictive ability of the model internally or externally

Full size image

Texture analysis

Texture analysis was the first method applied in image-processing research. A combined model based on clinical and texture features had C-indices of 0.83 for PFS and 0.90 for OS, which were higher than the corresponding values of the clinical model (0.68 and 0.78) [77].

Metabolic heterogeneity (MH) is a texture features in FDG PET/CT that is calculated from the AUC of the cumulative SUV-volume histogram corresponding to the lesion with the largest MTV [56, 76]. MH can quantify the variable coefficient of glucose uptake within the tumor and reflect the inhomogeneity of the tumor microenvironment [78]. Recent studies suggested that a high MH at DLBCL diagnosis predicted a worse outcome [51, 76, 79]. Patients with large MTV and MH had a 2-year PFS rate of 42% and experienced early relapse (median PFS 11.4 months) [79]. Senjo et al. used a model integrating MH and TMTV based on two independent DLBCL cohorts to stratify patients into three groups with significantly different outcomes (5-year OS: 90.4% vs 69.5% vs 34.8%, P < 0.001) [76]. These studies indicated that texture analysis might help to identify high-risk patients, enabling them to be offered intensive treatment at an early stage.

Radiomics

Radiomics conventionally constructs models by fixed algorithms using mass first-order and high-order features from images for clinical analysis. Conventional FDG PET/CT radiomics has been used in numerous lymphoma studies, and a systematic review showed that radiomics features could serve as diagnostic and prognostic indicators of lymphoma [80]. A model, which assessed by C-index and Akaike information criteria, based on WHO performance status, patient age, and radiomics provided better predictions of 2-year PFS and OS for DLBCL compared with the IPI risk score [72].

Radiomics is generally used to construct models based on combinations of features. An optimal model using clinical indicators and radiomics features predicted the 2-year time to progression of DLBCL with an AUC of 0.79 [68]. A wavelet transform model incorporating clinical indicators and FDG PET/CT radiomics yielded a higher AUC (0.75) than a model based solely on MTV (0.67) to predict 2-year event-free survival in patients with DLBCL [75]. A combined model of DLBCL progression based on metabolic metrics, clinical risk factors, and FDG PET/CT radiomics was superior to the single model and provided high C-indices for both the training set (PFS 0.825, OS 0.834) and validation set (PFS 0.831, OS 0.877) [74, 81]. A model combining BCL-6 and radiomics features dimensionally reduced using linear discriminant analysis had high predictive efficiency for DLBCL (AUC = 0.904, accuracy 90%, sensitivity 100%, specificity 80%) [82]. By comparing the time-dependent ROC curves, a nomogram including blood platelet count, sex, and radiomics score (Rad-scores) had been proven to provide a better recurrence risk assessment [83], and Zhao et al. also proved that a combination of different classifiers yielded a higher AUC for DLBCL prognosis than a single classifier [84].

Despite the promise of PET radiomics, some challenges still need to be addressed to improve model reliability and interpretability [80, 85]. However, radiomics models could be successfully used in DLBCL clinical settings if more robust prognostic models can be established using big data from multicenter studies.

Deep learning

Deep learning neural networks, especially convolutional neural networks (CNN), have been widely used to identify, segment and tumors, extract features and predict outcomes. [73, 86,87,88,89].

It’s crucial to identify early patients with bone marrow (BM) involvement since BM lymphoma invasion is a sign of advanced disease [90]. The BM lesions obtained from FDG PET/CT through the method of manual detection, radiomics or deep-learning combined with the results of bone marrow biopsy, could all improve the detection of BM involvement in patients with DLBCL and provide more accurate prognoses [26, 91,92,93,94]. For example, Jemaa et al. proved that patients with both positive biopsy and PET results, analyzed using a deep learning algorithm, have the worst prognosis compared to those with both negative results (2-year PFS: 62% vs. 72%) [26].

With the DLBCL patients as the training set, on the follicular lymphoma test set, a novel cascaded 2D to 3D CNN architecture produced a Dice Similarity Coefficient (DSC) of 0.886 and a voxel level sensitivity of 92.6% in identifying and segmenting tumors [87]. Weisman et al. implemented a 3D, multiresolution pathway CNN, DeepMedic to automatically detect lymph nodes involved in lymphoma and achieved a true-positive rate (TPR) of 85% [89]. Based on these predicted tumor mask, the automatic calculated TMTV could yield very precise estimates in Jemaa et al.’s study with Spearman’s correlations respectively of 0.97 compared with ground truth and predict the outcomes of DLBCL in Capobianco et al.’s study with 4-y OS rates were 90% and 74% for the low- and high-TMTV groups (optimal TMTV cutoffs: 148 cm3) [86, 87].

Haggstrom et al.’s ResNet34-based deep learning model could distinguish lymphoma patients, including DLBCL with and without hypermetabolic tumour sites, for binary classification (Deauville 1-3 vs 4-5), with AUC, accuracy, and sensitivity all exceeding 0.9 [95]. Jemaa et al.’s deep learning-based algorithm for automated metabolic response assessment in Lugano had strong prognostic value for outcomes. In three trials, there was a trend toward greater accuracy for risk of death than adjudicated radiologic responses (hazard ratio for end of treatment CMR of 0.123, 0.054, and 0.205 vs 0.226, 0.292, and 0.272, respectively) [96]. Deep learning for automated treatment response assessments in DLBCL would eventually change workflows and labor and resource allocation in clinical research and practice [97], as end-of-treatment PET response had been shown to be prognostic for OS [98,99,100].

Deep learning models have been applied in numerous diseases, including lymphoma, breast cancer, rectal cancer, and nasopharyngeal carcinoma [101,102,103]. Multiparametric models based on patient age, Ann Arbor stage, SUVmax, TMTV, and deep learning scores obtained from VGG19 and DenseNet121 networks were built to predict DLBCL prognosis, with C-indices 0.866 for PFS and 0.835 for OS, and were verified by C-index in external validation cohorts [104]. A deep learning model based on interim FDG PET/CT images showed good performance in a test cohort (AUC = 0.926) and external datasets (AUC = 0.925) for directing individualized clinical treatment of DLBCL patients [105].

Discussion

Comparing studies can summarize ways to boost research reliability. For multicenter radiomics studies, which are reliable than single-center studies, it is necessary to use the Combat method to assess the differences between various scanners. Most literature used ROC to rate models. Some of them implemented DeLong for AUC, assessing differences like sample size more rigorously. In addition to comparing the prognostic capabilities of models containing different indicators, there could be more articles comparing the models built using different machine learning methods, such as logistic regression with Least Absolute Shrinkage and Selection Operator (LASSO), ridge and elasticnet penalties, support vector machine and random forest.

However, the clinical adoption of artificial intelligence (AI) methods has been hindered by the lack of interpretability and generalizability, so increasing the interpretability of AI algorithms and creating a supervised deep-learning system for medical imaging based on a large, labeled dataset could gain the confidence of doctors and patients [106]. As the first step of the research, most studies used a single method to segment the lymphoma lesions. The heterogeneity in DLBCL lesion tracer uptake means a one-size-fits-all segmentation approach might not be ideal for all patients [68, 89]. Another key limitation is studies comparing human and AI diagnostics often lack real-world clinical context, relying solely on images without considering patient histories or additional data. This often increases the difficulty of the diagnostic task for the human reader [107].

Conclusion

DLBCL is a heterogeneous disease with series of baseline IPI-based scoring standards, which have been iteratively developed to address the heterogeneity of clinical outcomes. Molecular characteristics have been included to improve predictions of DLBCL progression and identify novel biological targets. Various baseline FDG PET/CT parameters, including MTV, TLG, and Dmax, have been used to construct machine learning models, but baseline radiomics and deep learning models that can predict the outcome of DLBCL remain in their infancy. Further development of AI technology will provide better predictive models based on big data from multicenter studies.

Data availability

No datasets were generated or analysed during the current study.

Abbreviations

aa-IPI:

Age-adjusted IPI

ABC:

Activated B-cell-like

AI:

Artificial intelligence

AUC:

Area under the curve

BM:

Bone marrow

CAR:

Chimeric antigen receptor

CCND2:

Cyclin D2

CMR:

Complete metabolic response

CNN:

Convolutional neural networks

CNS-IPI:

Central nervous system IPI

DA:

Domain adaptation

DCA:

Decision curve analysis

DLBCL:

Diffuse large B-cell lymphoma

ECOG:

Eastern Cooperative Oncology Group

FDG:

Fluorodeoxyglucose

GCB:

Germinal-center B-cell-like

IBPS:

Inflammation-Based Prognosis Score

IMPI:

International Metabolic Prognostic Index

IPI:

International Prognostic Index

IRF4:

Interferonregulatoryfactor4

K-M:

Kaplan-meier

MH:

Metabolic heterogeneity

MTV:

Metabolic tumor volume

MUM1:

Multiplemyelomaoncogene1

MV2:

Voxels detected by at least 2 methods

MV3:

Voxels detected by at least 3 methods

NCCN-IPI:

National Comprehensive Cancer Network IPI

OS:

Overall survival

PCA:

Principal component analysis

PET/CT:

Positron emission tomography computed tomography

PFS:

Progression-free survival

RF:

Random forest

ROC:

Receiver operating characteristic

R/R DLBCL:

Relapsed/ refractory DLBCL

R-IPI:

Revised IPI

SUV:

Standardized uptake value

SVM:

Support vector machine

TLG:

Total lesion glycolysis

WHO:

World Health Organization

References

  1. Melchardt T, Egle A, Greil R. How I treat diffuse large B-cell lymphoma. ESMO Open. 2023;8(1):100750.

  2. Sehn LH, Salles G. Diffuse Large B-Cell Lymphoma. N Engl J Med. 2021;384(9):842-58.

  3. Tavakkoli M, Barta SK. 2024 Update: advances in the risk stratification and management of large B-cell lymphoma. Am J Hematol. 2023;98(11):1791–805.

    Article  PubMed  Google Scholar 

  4. Dabrowska-Iwanicka AP, Nowakowski GS. DLBCL: who is high risk and how should treatment be optimized? Blood. 2023;11:2023020779.

  5. Munoz J, Deshpande A, Rimsza L, Nowakowski GS, Kurzrock R. Navigating between Scylla and Charybdis: a roadmap to do better than Pola-RCHP in DLBCL. Cancer Treat Rev. 2024;124:102691.

  6. Tilly H, Morschhauser F, Sehn LH, Friedberg JW, Trneny M, Sharman JP, et al. Polatuzumab vedotin in previously untreated diffuse large B-Cell lymphoma. N Engl J Med. 2022;386(4):351–63.

    Article  CAS  PubMed  Google Scholar 

  7. Susanibar-Adaniya S, Barta SK. 2021 update on diffuse large B cell lymphoma: a review of current data and potential applications on risk stratification and management. Am J Hematol. 2021;96(5):617–29.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Yan J, Yuan W, Zhang J, Li L, Zhang L, Zhang X, et al. Identification and validation of a prognostic prediction model in diffuse large B-Cell Lymphoma. Front Endocrinol (Lausanne). 2022;13:846357.

    Article  PubMed  Google Scholar 

  9. Rinehardt HN, Longo S, Gilbert R, Shoaf JN, Edwards WB, Kohanbash G, et al. Handheld PET probe for pediatric cancer surgery. Cancers (Basel). 2022;14(9):2221.

    Article  CAS  PubMed  Google Scholar 

  10. Reed JD, Masenge A, Buchner A, Omar F, Reynders D, Vorster M, et al. The utility of metabolic parameters on baseline F-18 FDG PET/CT in predicting treatment response and survival in paediatric and adolescent hodgkin lymphoma. J Clin Med. 2021;10(24):5979.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Mayerhoefer ME, Umutlu L, Schoder H. Functional imaging using radiomic features in assessment of lymphoma. Methods. 2021;188:105-11.

  12. Pomykala KL, Fendler WP, Vermesh O, Umutlu L, Herrmann K, Seifert R. Molecular Imaging of lymphoma: future directions and perspectives. Semin Nucl Med. 2023;53(3):449-56.

  13. Frood R, Burton C, Tsoumpas C, Frangi AF, Gleeson F, Patel C, et al. Baseline PET/CT imaging parameters for prediction of treatment outcome in Hodgkin and diffuse large B cell lymphoma: a systematic review. Eur J Nucl Med Mol Imaging. 2021;48(10):3198-220.

  14. Sehn LH, Berry B, Chhanabhai M, Fitzgerald C, Gill K, Hoskins P, et al. The revised International Prognostic Index (R-IPI) is a better predictor of outcome than the standard IPI for patients with diffuse large B-cell lymphoma treated with R-CHOP. Blood. 2007;109(5):1857-61.

  15. Ruppert AS, Dixon JG, Salles G, Wall A, Cunningham D, Poeschel V, et al. International prognostic indices in diffuse large B-cell lymphoma: a comparison of IPI, R-IPI, and NCCN-IPI. Blood. 2020;135(23):2041-8.

  16. Biccler J, Eloranta S, de Nully BP, Frederiksen H, Jerkeman M, Smedby KE, et al. Simplicity at the cost of predictive accuracy in diffuse large B-cell lymphoma: a critical assessment of the R-IPI, IPI, and NCCN-IPI. Cancer Med. 2018;7(1):114-22.

  17. Song JL, Wei XL, Zhang YK, Hao XX, Huang WM, Wei Q, et al. The prognostic value of the international prognostic index, the national comprehensive cancer network IPI and the age-adjusted IPI in diffuse large B cell lymphoma. Zhonghua Xue Ye Xue Za Zhi. 2018;39(9):739-44.

  18. Gleeson M, Counsell N, Cunningham D, Lawrie A, Clifton-Hadley L, Hawkes E, et al. Prognostic indices in diffuse large B-cell lymphoma in the rituximab era: an analysis of the UK National Cancer Research Institute R-CHOP 14 versus 21 phase 3 trial. Br J Haematol. 2021;192(6):1015-9.

  19. Zhou Z, Sehn LH, Rademaker AW, Gordon LI, Lacasce AS, Crosby-Thompson A, et al. An enhanced International Prognostic Index (NCCN-IPI) for patients with diffuse large B-cell lymphoma treated in the rituximab era. Blood. 2014;123(6):837-42.

  20. Jelicic J, Juul-Jensen K, Bukumiric Z, Roost Clausen M, Ludvigsen Al-Mashhadi A, Pedersen RS, et al. Prognostic indices in diffuse large B-cell lymphoma: a population-based comparison and validation study of multiple models. Blood Cancer J. 2023;13(1):157.

    Article  PubMed  PubMed Central  Google Scholar 

  21. Ellin F, Maurer MJ, Srour L, Farooq U, Jerkeman M, Connors JM, et al. Comparison of the NCCN-IPI, the IPI and PIT scores as prognostic tools in peripheral T-cell lymphomas. Br J Haematol. 2019;186(3):e24-7.

  22. Yhim HY, Park Y, Han YH, Kim S, Kang SR, Moon JH, et al. A risk stratification model for nodal peripheral T-cell lymphomas based on the NCCN-IPI and posttreatment Deauville score. Eur J Nucl Med Mol Imaging. 2018;45(13):2274-84.

  23. Schmitz N, Zeynalova S, Nickelsen M, Kansara R, Villa D, Sehn LH, et al. CNS international prognostic index: a risk model for CNS relapse in patients with diffuse large b-cell lymphoma treated with R-CHOP. J Clin Oncol. 2016;34(26):3150-6.

  24. Fischer T, Zing NP, Fortier SC, Schmidt J, Silveira TB, Chiattone CS. Application of the Central Nervous System International Prognostic Index (CNS-IPI) score in daily practice: a retrospective analysis apart from the clinical trial at two centers in Brazil. Hematol Transfus Cell Ther. 2024;46(2):137-45.

  25. Solis-Armenta R, Cacho-Diaz B, Gutierrez-Hernandez O, Candelaria-Hernandez M. Central nervous system international prognostic index impacts overall survival in diffuse large b-cell lymphoma treated with r-chop in a third level cancer center from mexico: a survey of 642 patients. Rev Invest Clin. 2021;73(4):231-7.

  26. Jemaa S, Paulson JN, Hutchings M, Kostakoglu L, Trotman J, Tracy S, et al. Full automation of total metabolic tumor volume from FDG-PET/CT in DLBCL for baseline risk assessments. Cancer Imaging. 2022;22(1):39.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Klanova M, Sehn LH, Bence-Bruckler I, Cavallo F, Jin J, Martelli M, et al. Integration of cell of origin into the clinical CNS International Prognostic Index improves CNS relapse prediction in DLBCL. Blood. 2019;133(9):919-26.

  28. Liu Y, Sheng L, Hua H, Zhou J, Zhao Y, Wang B. A Novel and Validated Inflammation-Based Prognosis Score (IBPS) predicts outcomes in patients with diffuse large b-cell lymphoma. Cancer Manag Res. 2023;15:651-66.

  29. Maurer MJ, Jakobsen LH, Mwangi R, Schmitz N, Farooq U, Flowers CR, et al. Relapsed/Refractory International Prognostic Index (R/R-IPI): an international prognostic calculator for relapsed/refractory diffuse large B-cell lymphoma. Am J Hematol. 2021;96(5):599-605.

  30. Kim J, Cho J, Yoon SE, Kim WS, Kim SJ. Efficacy of salvage treatments in relapsed or refractory diffuse large b-cell lymphoma including chimeric antigen receptor t-cell therapy: a systematic review and meta-analysis. Cancer Res Treat. 2023;55(3):1031-47.

  31. Bento L, Diaz-Lopez A, Barranco G, Martin-Moreno AM, Baile M, Martin A, et al. New prognosis score including absolute lymphocyte/monocyte ratio, red blood cell distribution width and beta-2 microglobulin in patients with diffuse large B-cell lymphoma treated with R-CHOP: Spanish Lymphoma Group Experience (GELTAMO). Br J Haematol. 2020;188(6):888-97.

  32. Li M, Xia H, Zheng H, Li Y, Liu J, Hu L, et al. Red blood cell distribution width and platelet counts are independent prognostic factors and improve the predictive ability of IPI score in diffuse large B-cell lymphoma patients. BMC Cancer. 2019;19(1):1084.

    Article  PubMed  PubMed Central  Google Scholar 

  33. Gao R, Liang JH, Wang L, Zhu HY, Wu W, Cao L, et al. Low serum cholesterol levels predict inferior prognosis and improve NCCN-IPI scoring in diffuse large B cell lymphoma. Int J Cancer. 2018;143(8):1884-95.

  34. Tavakkoli M, Barta SK. 2024 Update: Advances in the risk stratification and management of large B-cell lymphoma. Am J Hematol. 2023;98(11):1791-805.

  35. Weber T, Schmitz R. Molecular subgroups of diffuse large B cell lymphoma: biology and implications for clinical practice. Curr Oncol Rep. 2022;24(1):13-21.

  36. Painter D, Barrans S, Lacy S, Smith A, Crouch S, Westhead D, et al. Cell-of-origin in diffuse large B-cell lymphoma: findings from the UK’s population-based Haematological Malignancy Research Network. Br J Haematol. 2019;185(4):781-4.

  37. Stegemann M, Denker S, Schmitt CA. DLBCL 1L-what to expect beyond R-CHOP? Cancers (Basel). 2022;14(6):1453.

    Article  CAS  PubMed  Google Scholar 

  38. Riedell PA, Smith SM. Double hit and double expressors in lymphoma: definition and treatment. Cancer. 2018;124(24):4622-32.

  39. Li W, Gupta SK, Han W, Kundson RA, Nelson S, Knutson D, et al. Targeting MYC activity in double-hit lymphoma with MYC and BCL2 and/or BCL6 rearrangements with epigenetic bromodomain inhibitors. J Hematol Oncol. 2019;12(1):73.

    Article  PubMed  PubMed Central  Google Scholar 

  40. Campo E, Jaffe ES, Cook JR, Quintanilla-Martinez L, Swerdlow SH, Anderson KC, et al. The international consensus classification of mature lymphoid neoplasms: a report from the clinical advisory committee. Blood. 2022;140(11):1229-53.

  41. El-Sharkawi D, Sud A, Prodger C, Khwaja J, Shotton R, Hanley B, et al. A retrospective study of MYC rearranged diffuse large B-cell lymphoma in the context of the new WHO and ICC classifications. Blood Cancer J. 2023;13(1):54.

    Article  PubMed  PubMed Central  Google Scholar 

  42. He J, Chen Z, Xue Q, Sun P, Wang Y, Zhu C, et al. Identification of molecular subtypes and a novel prognostic model of diffuse large B-cell lymphoma based on a metabolism-associated gene signature. J Transl Med. 2022;20(1):186.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  43. Zhang MC, Tian S, Fu D, Wang L, Cheng S, Yi HM, et al. Genetic subtype-guided immunochemotherapy in diffuse large B cell lymphoma: the randomized GUIDANCE-01 trial. Cancer Cell. 2023;41(10):1705-16 e5.

    Article  CAS  PubMed  Google Scholar 

  44. Wang D, Zhang Y, Che YQ. CCND2 mRNA expression is correlated with R-CHOP treatment efficacy and prognosis in patients with ABC-DLBCL. Front Oncol. 2020;10:1180.

    Article  PubMed  PubMed Central  Google Scholar 

  45. Ning N, Zhang S, Wu Q, Li X, Kuang D, Duan Y, et al. Inhibition of acylglycerol kinase sensitizes DLBCL to venetoclax via upregulation of FOXO1-mediated BCL-2 expression. Theranostics. 2022;12(12):5537-50.

  46. El-Galaly TC, Villa D, Gormsen LC, Baech J, Lo A, Cheah CY. FDG-PET/CT in the management of lymphomas: current status and future directions. J Intern Med. 2018;284(4):358-76.

  47. Chantepie S, Hovhannisyan N, Guillouet S, Pelage JP, Ibazizene M, Bodet-Milin C, et al. (18)F-Fludarabine PET for lymphoma imaging: first-in-humans study on DLBCL and CLL patients. J Nucl Med. 2018;59(9):1380-5.

  48. Kostakoglu L, Mattiello F, Martelli M, Sehn LH, Belada D, Ghiggi C, et al. Total metabolic tumor volume as a survival predictor for patients with diffuse large B-cell lymphoma in the GOYA study. Haematologica. 2022;107(7):1633-42.

  49. Zhang YY, Song L, Zhao MX, Hu K. A better prediction of progression-free survival in diffuse large B-cell lymphoma by a prognostic model consisting of baseline TLG and %DeltaSUV(max). Cancer Med. 2019;8(11):5137-47.

  50. Mettler J, Muller H, Voltin CA, Baues C, Klaeser B, Moccia A, et al. Metabolic tumour volume for response prediction in advanced-stage hodgkin lymphoma. J Nucl Med. 2018;60(2):207-11.

  51. Ceriani L, Gritti G, Cascione L, Pirosa MC, Polino A, Ruberto T, et al. SAKK38/07 study: integration of baseline metabolic heterogeneity and metabolic tumor volume in DLBCL prognostic model. Blood Adv. 2020;4(6):1082-92.

  52. Barrington SF, Zwezerijnen B, de Vet HCW, Heymans MW, Mikhaeel NG, Burggraaff CN, et al. Automated segmentation of baseline metabolic total tumor burden in diffuse large B-Cell lymphoma: which method is most successful? A study on Behalf of the PETRA Consortium. J Nucl Med. 2021;62(3):332-7.

  53. El-Galaly TC, Villa D, Cheah CY, Gormsen LC. Pre-treatment total metabolic tumour volumes in lymphoma: does quantity matter? Br J Haematol. 2022;197(2):139-55.

  54. Zwezerijnen GJC, Eertink JJ, Burggraaff CN, Wiegers SE, Shaban E, Pieplenbosch S, et al. Interobserver agreement on automated metabolic tumor volume measurements of deauville score 4 and 5 lesions at interim (18)F-FDG PET in diffuse large B-Cell lymphoma. J Nucl Med. 2021;62(11):1531-6.

  55. Czibor S, Carr R, Redondo F, Auewarakul CU, Cerci JJ, Paez D, et al. Prognostic parameters on baseline and interim [18F]FDG-PET/computed tomography in diffuse large B-cell lymphoma patients. Nucl Med Commun. 2023;44(4):291-301.

  56. Ceriani L, Milan L, Martelli M, Ferreri AJM, Cascione L, Zinzani PL, et al. Metabolic heterogeneity on baseline 18FDG-PET/CT scan is a predictor of outcome in primary mediastinal B-cell lymphoma. Blood. 2018;132(2):179-86.

  57. Vercellino L, Cottereau AS, Casasnovas O, Tilly H, Feugier P, Chartier L, et al. High total metabolic tumor volume at baseline predicts survival independent of response to therapy. Blood. 2020;135(16):1396-405.

  58. Cottereau AS, Versari A, Loft A, Casasnovas O, Bellei M, Ricci R, et al. Prognostic value of baseline metabolic tumor volume in early-stage Hodgkin lymphoma in the standard arm of the H10 trial. Blood. 2018;131(13):1456-63.

  59. Thieblemont C, Chartier L, Duhrsen U, Vitolo U, Barrington SF, Zaucha JM, et al. A tumor volume and performance status model to predict outcome before treatment in diffuse large B-cell lymphoma. Blood Adv. 2022;6(23):5995-6004.

  60. Vercellino L, Di Blasi R, Kanoun S, Tessoulin B, Rossi C, D’Aveni-Piney M, et al. Predictive factors of early progression after CAR T-cell therapy in relapsed/refractory diffuse large B-cell lymphoma. Blood Adv. 2020;4(22):5607-15.

  61. Mikhaeel NG, Heymans MW, Eertink JJ, de Vet HCW, Boellaard R, Duhrsen U, et al. Proposed new dynamic prognostic index for diffuse large B-Cell lymphoma: international metabolic prognostic index. J Clin Oncol. 2022;40(21):2352-60.

  62. Winkelmann M, Blumenberg V, Rejeski K, Bucklein VL, Ruzicka M, Unterrainer M, et al. Prognostic value of the International Metabolic Prognostic Index for lymphoma patients receiving chimeric antigen receptor T-cell therapy. Eur J Nucl Med Mol Imaging. 2023;50(5):1406-13.

  63. Alderuccio JP, Reis IM, Hamadani M, Nachiappan M, Leslom S, Kahl BS, et al. PET/CT biomarkers enable risk stratification of patients with relapsed/refractory diffuse large B-cell lymphoma enrolled in the LOTIS-2 clinical trial. Clin Cancer Res. 2024;30(1):139-49.

  64. Zhao P, Yu T, Pan Z. Prognostic value of the baseline 18F-FDG PET/CT metabolic tumour volume (MTV) and further stratification in low-intermediate (L-I) and high-intermediate (H-I) risk NCCNIPI subgroup by MTV in DLBCL MTV predict prognosis in DLBCL. Ann Nucl Med. 2021;35(1):24-30.

  65. Ma SY, Tian XP, Cai J, Su N, Fang Y, Zhang YC, et al. A prognostic immune risk score for diffuse large B-cell lymphoma. Br J Haematol. 2021;194(1):111-9.

  66. Xu H, Ma J, Yang G, Xiao S, Li W, Sun Y, et al. Prognostic value of metabolic tumor volume and lesion dissemination from baseline PET/CT in patients with diffuse large B-cell lymphoma: further risk stratification of the group with low-risk and high-risk NCCN-IPI. Eur J Radiol. 2023;163:110798.

    Article  PubMed  Google Scholar 

  67. Cottereau AS, Nioche C, Dirand AS, Clerc J, Morschhauser F, Casasnovas O, et al. (18)F-FDG PET dissemination features in diffuse large B-cell lymphoma are predictive of outcome. J Nucl Med. 2020;61(1):40-5.

  68. Eertink JJ, van de Brug T, Wiegers SE, Zwezerijnen GJC, Pfaehler EAG, Lugtenburg PJ, et al. (18)F-FDG PET baseline radiomics features improve the prediction of treatment outcome in diffuse large B-cell lymphoma. Eur J Nucl Med Mol Imaging. 2022;49(3):932-42.

  69. Jo JH, Chung HW, Kim SY, Lee MH, So Y. FDG PET/CT maximum tumor dissemination to predict recurrence in patients with diffuse large B-Cell Lymphoma. Nucl Med Mol Imaging. 2023;57(1):26-33.

  70. Eertink JJ, Zwezerijnen GJC, Cysouw MCF, Wiegers SE, Pfaehler EAG, Lugtenburg PJ, et al. Comparing lesion and feature selections to predict progression in newly diagnosed DLBCL patients with FDG PET/CT radiomics features. Eur J Nucl Med Mol Imaging. 2022;49(13):4642-51.

  71. Cottereau AS, Meignan M, Nioche C, Capobianco N, Clerc J, Chartier L, et al. Risk stratification in diffuse large B-cell lymphoma using lesion dissemination and metabolic tumor burden calculated from baseline PET/CT(dagger). Ann Oncol. 2021;32(3):404-11.

  72. Eertink JJ, Zwezerijnen GJC, Heymans MW, Pieplenbosch S, Wiegers SE, Duhrsen U, et al. Baseline PET radiomics outperforms the IPI risk score for prediction of outcome in diffuse large B-cell lymphoma. Blood. 2023;141(25):3055-64.

  73. Avanzo M, Wei L, Stancanello J, Vallieres M, Rao A, Morin O, et al. Machine and deep learning methods for radiomics. Med Phys. 2020;47(5):e185-202.

  74. Jiang C, Huang X, Li A, Teng Y, Ding C, Chen J, et al. Radiomics signature from [(18)F]FDG PET images for prognosis predication of primary gastrointestinal diffuse large B cell lymphoma. Eur Radiol. 2022;32(8):5730-41.

  75. Frood R, Clark M, Burton C, Tsoumpas C, Frangi AF, Gleeson F, et al. Discovery of pre-treatment FDG PET/CT-derived radiomics-based models for predicting outcome in diffuse large B-Cell lymphoma. Cancers (Basel). 2022;14(7):1711.

    Article  CAS  PubMed  Google Scholar 

  76. Senjo H, Hirata K, Izumiyama K, Minauchi K, Tsukamoto E, Itoh K, et al. High metabolic heterogeneity on baseline 18FDG-PET/CT scan as a poor prognostic factor for newly diagnosed diffuse large B-cell lymphoma. Blood Adv. 2020;4(10):2286-96.

  77. Travaini LL, Botta F, Derenzini E, Lo Presti G, Ferrari ME, Airo Farulla LS, et al. [(18) F]-FDG PET radiomic model as prognostic biomarker in diffuse large B-cell lymphoma. Hematol Oncol. 2023;41(4):674-82.

  78. van Velden FH, Cheebsumon P, Yaqub M, Smit EF, Hoekstra OS, Lammertsma AA, et al. Evaluation of a cumulative SUV-volume histogram method for parameterizing heterogeneous intratumoural FDG uptake in non-small cell lung cancer PET studies. Eur J Nucl Med Mol Imaging. 2011;38(9):1636-47.

  79. Genta S, Ghilardi G, Cascione L, Juskevicius D, Tzankov A, Schar S, et al. Integration of baseline metabolic parameters and mutational profiles predicts long-term response to first-line therapy in DLBCL patients: a post hoc analysis of the SAKK38/07 study. Cancers (Basel). 2022;14(4):1018.

    Article  CAS  PubMed  Google Scholar 

  80. Wang H, Zhou Y, Li L, Hou W, Ma X, Tian R. Current status and quality of radiomics studies in lymphoma: a systematic review. Eur Radiol. 2020;30(11):6228-40.

  81. Jiang C, Li A, Teng Y, Huang X, Ding C, Chen J, et al. Optimal PET-based radiomic signature construction based on the cross-combination method for predicting the survival of patients with diffuse large B-cell lymphoma. Eur J Nucl Med Mol Imaging. 2022;49(8):2902–16.

    Article  PubMed  Google Scholar 

  82. Ferrer-Lores B, Lozano J, Fuster-Matanzo A, Mayorga-Ruiz I, Moreno-Ruiz P, Bellvís F. Prognostic value of genetic alterations and 18F-FDG PET/CT imaging features in diffuse large B cell lymphoma. Am J Cancer Res 2023;13(2):509-25.

  83. Li M, Yao H, Zhang P, Zhang L, Liu W, Jiang Z, et al. Development and validation of a [(18)F]FDG PET/CT-based radiomics nomogram to predict the prognostic risk of pretreatment diffuse large B cell lymphoma patients. Eur Radiol. 2023;33(5):3354-65.

  84. Zhao S, Wang J, Jin C, Zhang X, Xue C, Zhou R, et al. Stacking ensemble learning-based [(18)F]FDG PET radiomics for outcome prediction in diffuse large B-Cell lymphoma. J Nucl Med. 2023;64(10):1603-9.

  85. Pineiro-Fiel M, Moscoso A, Pubul V, Ruibal A, Silva-Rodriguez J, Aguiar P. A systematic review of PET textural analysis and radiomics in cancer. Diagnostics (Basel). 2021;11(2):380.

    Article  CAS  PubMed  Google Scholar 

  86. Capobianco N, Meignan M, Cottereau AS, Vercellino L, Sibille L, Spottiswoode B, et al. Deep-learning (18)F-FDG uptake classification enables total metabolic tumor volume estimation in diffuse large B-Cell lymphoma. J Nucl Med. 2021;62(1):30-6.

  87. Jemaa S, Fredrickson J, Carano RAD, Nielsen T, de Crespigny A, Bengtsson T. Tumor segmentation and feature extraction from whole-body FDG-PET/CT using cascaded 2D and 3D convolutional neural networks. J Digit Imaging. 2020;33(4):888-94.

  88. Sibille L, Seifert R, Avramovic N, Vehren T, Spottiswoode B, Zuehlsdorff S, et al. (18)F-FDG PET/CT Uptake classification in lymphoma and lung cancer by using deep convolutional neural networks. Radiology. 2020;294(2):445-52.

  89. Weisman AJ, Kieler MW, Perlman SB, Hutchings M, Jeraj R, Kostakoglu L, et al. Convolutional neural networks for automated PET/CT detection of diseased lymph node burden in patients with lymphoma. Radiol Artif Intell. 2020;2(5):e200016.

    Article  PubMed  PubMed Central  Google Scholar 

  90. Yao Z, Deng L, Xu-Monette ZY, Manyam GC, Jain P, Tzankov A, et al. Concordant bone marrow involvement of diffuse large B-cell lymphoma represents a distinct clinical and biological entity in the era of immunotherapy. Leukemia. 2018;32(2):353-63.

  91. Han EJ, O JH, Yoon H, Ha S, Yoo IR, Min JW, et al. Comparison of FDG PET/CT and bone marrow biopsy results in patients with diffuse large B cell lymphoma with subgroup analysis of PET radiomics. Diagnostics (Basel). 2022;12(1):222.

    Article  CAS  PubMed  Google Scholar 

  92. Cerci JJ, Gyorke T, Fanti S, Paez D, Meneghetti JC, Redondo F, et al. Combined PET and biopsy evidence of marrow involvement improves prognostic prediction in diffuse large B-cell lymphoma. J Nucl Med. 2014;55(10):1591-7.

  93. St-Pierre F, Broski SM, LaPlant BR, Ristow K, Maurer MJ, Macon WR, et al. Detection of extranodal and spleen involvement by FDG-PET imaging predicts adverse survival in untreated follicular lymphoma. Am J Hematol. 2019;94(7):786-93.

  94. Weiler-Sagie M, Kagna O, Dann EJ, Ben-Barak A, Israel O. Characterizing bone marrow involvement in Hodgkin’s lymphoma by FDG-PET/CT. Eur J Nucl Med Mol Imaging. 2014;41(6):1133-40.

  95. Häggström I, Leithner D, Alvén J, Campanella G, Abusamra M, Zhang H, et al. Deep learning for [18F]fluorodeoxyglucose-PET-CT classification in patients with lymphoma: a dual-centre retrospective analysis. Lancet Digit Health. 2024;6(2):e114-25.

    Article  PubMed  Google Scholar 

  96. Jemaa S, Ounadjela S, Wang X, El-Galaly TC, Kostakoglu L, Knapp A, et al. Automated Lugano metabolic response assessment in 18F-fluorodeoxyglucose–avid non-hodgkin lymphoma with deep learning on 18F-fluorodeoxyglucose–positron emission tomography. J Clin Oncol. 2024;42(25):2966–77.

  97. Schöder H. Machine learning for automated interpretation of fluorodeoxyglucose-positron emission tomography scans in lymphoma. J Clin Oncol. 2024;42(25):2945–8.

  98. Cheson BD, Fisher RI, Barrington SF, Cavalli F, Schwartz LH, Zucca E, et al. Recommendations for initial evaluation, staging, and response assessment of hodgkin and non-hodgkin lymphoma: the Lugano classification. J Clin Oncol. 2014;32(27):3059–67.

  99. Kostakoglu L, Martelli M, Sehn LH, Belada D, Carella A-M, Chua N, et al. End-of-treatment PET/CT predicts PFS and OS in DLBCL after first-line treatment: results from GOYA. Blood Adv. 2021;5(5):1283–90.

  100. Trotman J, Barrington SF, Belada D, Meignan M, MacEwan R, Owen C, et al. Prognostic value of end-of-induction PET response after first-line immunochemotherapy for follicular lymphoma (GALLIUM): secondary analysis of a randomised, phase 3 trial. Lancet Oncol. 2018;19(11):1530–42.

  101. Zhong L, Dong D, Fang X, Zhang F, Zhang N, Zhang L, et al. A deep learning-based radiomic nomogram for prognosis and treatment decision in advanced nasopharyngeal carcinoma: a multicentre study. EBioMedicine. 2021;70:103522.

    Article  PubMed  PubMed Central  Google Scholar 

  102. Gu J, Tong T, Xu D, Cheng F, Fang C, He C, et al. Deep learning radiomics of ultrasonography for comprehensively predicting tumor and axillary lymph node status after neoadjuvant chemotherapy in breast cancer patients: a multicenter study. Cancer. 2023;129(3):356–66.

  103. Liu X, Zhang D, Liu Z, Li Z, Xie P, Sun K, et al. Deep learning radiomics-based prediction of distant metastasis in patients with locally advanced rectal cancer after neoadjuvant chemoradiotherapy: a multicentre study. EBioMedicine. 2021;69:103442.

    Article  PubMed  PubMed Central  Google Scholar 

  104. Jiang C, Qian C, Jiang Z, Teng Y, Lai R, Sun Y, et al. Robust deep learning-based PET prognostic imaging biomarker for DLBCL patients: a multicenter study. Eur J Nucl Med Mol Imaging. 2023;50(13):3949-60.

  105. Yuan C, Shi Q, Huang X, Wang L, He Y, Li B, et al. Multimodal deep learning model on interim [(18)F]FDG PET/CT for predicting primary treatment failure in diffuse large B-cell lymphoma. Eur Radiol. 2023;33(1):77–88.

  106. Jiang Y, Yang M, Wang S, Li X, Sun Y. Emerging role of deep learning-based artificial intelligence in tumor pathology. Cancer Commun (Lond). 2020;40(4):154–66.

  107. Esteva A, Robicquet A, Ramsundar B, Kuleshov V, DePristo M, Chou K, et al. A guide to deep learning in healthcare. Nat Med. 2019;25(1):24–9.

Download references

Acknowledgements

Thanks for the support from the teachers and classmates in the department of nuclear medicine of The Affiliated Huaian No.1 People’s Hospital of Nanjing Medical University.

Thanks for the support from the department of hematology of The Affiliated Huaian No.1 People’s Hospital of Nanjing Medical University.

Funding

This study was funded by Huai’an Science and Technology Project (grant no. HAB202017 to Weijing Tao), and the innovation key talents Project of the hospital (grant no. ZC202208 to Weijing Tao).

Author information

Authors and Affiliations

  1. Department of Nuclear Medicine, The Affiliated Huaian No.1 People’s Hospital of Nanjing Medical University, Huai’an, Jiangsu, China

    Zhuxu Sun, Tianshuo Yang, Luyi Cheng & Weijing Tao

  2. Department of Nuclear Medicine, The First Affiliated Hospital of Nanjing Medical University, Nanjing, Jiangsu, China

    Chongyang Ding

  3. Department of Hematology, The Affiliated Huaian No.1 People’s Hospital of Nanjing Medical University, Huai’an, Jiangsu, China

    Yuye Shi

  4. State Key Laboratory of Medicinal Chemical Biology, College of Pharmacy, Tianjin Central Hospital of Gynecology Obstetrics, Tianjin Key Laboratory of Human Development and Reproductive Regulation, Nankai University, Tianjin, China

    Qingshen Jia

Authors
  1. Zhuxu Sun
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Tianshuo Yang
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Chongyang Ding
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Yuye Shi
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Luyi Cheng
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Qingshen Jia
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Weijing Tao
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

TY, LC and QJ collected the literature. CD and YS offered clinical guidance for the review. WT organized thoughts and revised the review. All authors read and approved the final version of the manuscript.

Corresponding author

Correspondence to Weijing Tao.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Sun, Z., Yang, T., Ding, C. et al. Clinical scoring systems, molecular subtypes and baseline [18F]FDG PET/CT image analysis for prognosis of diffuse large B-cell lymphoma. Cancer Imaging 24, 168 (2024). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40644-024-00810-8

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40644-024-00810-8

Keywords

Cancer Imaging

ISSN: 1470-7330

Contact us