Background This multicenter phase II trial aimed to evaluate the orelabrutinib, fludarabine, cyclophosphamide, and obinutuzumab (OFCG) regimen as first-line therapy for patients with chronic lymphocytic leukemia (CLL)/small lymphocytic lymphoma without restriction by TP53 aberrations [del(17p) and/or TP53 mutation] and IGHV status. Methods Eligible patients received OFCG (orelabrutinib, orally 150 mg/day; fludarabine, intravenously 25 mg/m2/day; cyclophosphamide, intravenously 250 mg/m2/day; obinutuzumab, intravenously 1000 mg), with treatment duration and modifications based on prespecified minimal residual disease (MRD) criteria. The primary endpoint was the rate of undetectable MRD in bone marrow (BM-uMRD) after 6 cycles by flow cytometry (10−4 sensitivity). The key secondary endpoints included the rate of BM-uMRD4 at the end of cycles 3, 12, MRD in peripheral blood (PB-uMRD4) at the end of cycles 6, and complete response rate (CRR); exploratory endpoints included the rate of BM-uMRD6 (10−6 sensitivity) at the end of cycles 3, 6, and 12. Results Among 25 enrolled patients, 23 completed 3 cycles of OFCG therapy; of these, 13 patients achieved BM-uMRD4. After 6 cycles, 76.0 % (19/25) achieved BM-uMRD4, and 80 % (20/25) of patients achieved PB-uMRD4. CRR with BM-uMRD4 was achieved by 20.0 % (5/25), 52.0 % (13/25), and 64.0 % (16/25) of patients after 3, 6, and 12 cycles, respectively. Significant differences in BM-uMRD6 at C7D0 (odds ratio [OR], 0.03 [95 % confidence interval [CI]: 0, 0.32]; P < 0.01) and C13D0 (OR, 0.16 [95 % CI: 0.03, 0.94]; P < 0.05) were observed between IGHV-mutated (n = 10) and unmutated (n = 15) patients. Notably, 20 patients discontinued all the drugs at the data cutoff date. No treatment-related deaths occurred, and the most frequent grade 3–4 adverse effects (AEs) were neutropenia (20/25, 80.0 %), thrombocytopenia (15/25, 60.0 %), and infection (7/25, 28.0 %). Conclusion The study meet the primary endpoint, showing the potential efficacy and acceptable safety profile of the OFCG regimen as a first-line treatment for CLL. Trial registration: NCT05322733.

Fever of unknown origin (FUO) remains diagnostically challenging because of heterogeneous causes, non-specific clinical manifestations, and overlapping imaging findings. We developed and validated FUO-PETMamba, a PET maximum-intensity-projection (MIP)-based artificial intelligence framework for AI-assisted aetiological classification of FUO. This retrospective multicentre study included 681 patients with FUO who underwent baseline [18 F]FDG PET/CT, comprising one development cohort (n = 355) and two independent external validation cohorts (n = 195 and n = 131). FUO-PETMamba is a weakly supervised framework analysing PET MIP images generated from PET data. Model performance was assessed by discrimination, calibration, decision curve analysis (DCA). Attention-based visual explanations were generated using attention mechanisms and gradient-based activation mapping. A reader study assessed the potential assistive effect of model-predicted probabilities on physicians with different PET/CT experience. In the development cohort, FUO-PETMamba achieved AUCs of 0.838 for malignancy, 0.851 for infection, 0.914 for autoimmune disease, and 0.788 for miscellaneous causes. Corresponding AUCs were 0.809, 0.815, 0.805, and 0.849 in external validation cohort 1, and 0.808, 0.772, 0.701, and 0.927 in external validation cohort 2, respectively. Calibration and decision curve analysis suggested potential clinical benefit for the major aetiological categories, although performance varied across cohorts and classes. The miscellaneous category should be interpreted cautiously because of limited case numbers and low positive predictive value and F1 scores. In the reader study, AI assistance improved diagnostic accuracy for junior and intermediate physicians, whereas changes in senior-physician performance were variable. Post hoc attention-based visualisations highlighted clinically plausible hypermetabolic patterns and served as qualitative explanatory aids. FUO-PETMamba provides a PET MIP-based AI-assisted diagnostic support framework for aetiological classification of FUO across multicentre cohorts and may help reduce experience-dependent diagnostic variability after prospective validation.

Abstract Background Mantle cell lymphoma (MCL) is a rare, biologically heterogeneous B-cell malignancy with highly variable outcomes. Existing prognostic tools are suboptimal. We developed an interpretable deep learning framework integrating baseline [18F]FDG PET/CT and electronic health record (EHR) data for individualized risk stratification. Methods In this multicenter study, 187 treatment-naïve MCL patients were analyzed. A mixture-of-experts (MoE) fusion network integrated multimodal representations from PET/CT and EHR data. Expert modules comprising vision encoders, radiomics extractors, and a medical language model were integrated through an attention-based gating mechanism to construct multimodal radiomic signatures (R-signatures) predictive of progression-free survival (PFS) and overall survival (OS). R-signatures were validated and incorporated with clinical and metabolic factors into multiparametric models. Deep learning model interpretability was evaluated using attention visualization, expert-level contributions and pathologic correlation. Results R-signatures robustly discriminated relapse (AUC = 0.893 training, 0.755 validation) and death (AUC = 0.804 and 0.844), and independently predicted adverse outcomes (PFS: HR = 27.70, P < 0.001; OS: HR = 6.86, P = 0.001). Multiparametric models integrating R-signatures with total lesion glycolysis, β2-microglobulin, WBC, and Ki-67 outperformed conventional indices (C-indices: PFS 0.892 training, 0.781 validation; OS 0.877 training, 0.862 validation). Time-dependent ROC analyses consistently showed AUCs approaching or exceeding 0.800. Calibration and decision curve analyses confirmed excellent agreement and superior clinical net benefit. Attention maps localized high-weighted regions to hypermetabolic tumor areas, with higher R-signature values in blastoid and pleomorphic variants versus classical histology (P = 0.028 and P = 0.010). Conclusions This interpretable PET/CT-EHR fusion framework substantially improves prognostic precision in MCL, providing a noninvasive, clinically translatable tool for risk-adapted management.

To develop and validate a prognostic imaging biomarker derived from baseline [¹⁸F]FDG PET/CT using tabular deep learning for prediction of progression of disease within 24 months (POD24) and survival risk stratification in patients with follicular lymphoma (FL). This retrospective multicenter study included 309 patients with newly diagnosed FL (grades 1-3a) from five independent medical centers. Tumor volumes segmented from baseline [¹⁸F]FDG PET and CT images were used to extract high-throughput radiomic features. Five conventional machine learning algorithms and four advanced tabular deep learning models were developed and compared. The predictive output of the GAMformer model was defined as the deep learning score (DLS). The DLS was integrated with clinical variables and PET metabolic parameters to construct a multiparametric model in the training cohort. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis, and further validated in validation cohort. During a median follow-up of 44 months, POD24 occurred in 55 patients. The DLS demonstrated strong predictive performance for POD24 (training AUC = 0.857; validation AUC = 0.753). The multiparametric model further improved discrimination, achieving AUCs of 0.882 in the training cohort and 0.797 in the validation cohort, outperforming FLIPI, FLIPI-2, and PRIMA-PI. Calibration showed good agreement, and decision curve analysis indicated higher net clinical benefit. The DLS stratified survival risk (P < 0.05) and remained predictive of survival and POD24 across histologic grades. The DLS derived from baseline [¹⁸F]FDG PET/CT enables POD24 prediction and accurate survival risk stratification in FL, supporting its potential role in precision management.

Accurate grading of follicular lymphoma (FL) is crucial for personalized treatment, but biopsy-based histopathology is invasive and limited by observer variability. To address these limits, we present an artificial intelligence framework for physician-guided assisted FL grading using PET-CT imaging. Our approach integrates an enhanced dual-discriminator conditional GAN (DDCGAN) featuring similarity and chrominance constraints to generate task-specific fused images with preserved metabolic-structural cues. Furthermore, a Bayesian ResNet is introduced to explicitly model predictive uncertainty, effectively resolving classification ambiguity between adjacent FL Grades I and II. Rigorous evaluation on a multi-center dataset of 837 patients, including FL and diffuse large B-cell lymphoma (DLBCL), proves that our framework delivers superior generalizability. It achieves an accuracy of 0.871, precision of 0.875, and macro-F1 of 0.816, outperforming single-modality and existing state-of-the-art fusion models. Ultimately, this task-oriented image fusion and uncertainty-aware framework offers a highly practical, non-invasive decision-support tool to support scalable clinical decision-making in hospital workflows.

While involved-site radiotherapy (ISRT) is the standard first-line treatment in localized non-gastric mucosa-associated lymphoid tissue (MALT) lymphoma, the cumulative risk of distant relapse poses a persistent clinical challenge. We conducted a prospective phase II trial evaluating rituximab with 24 Gy ISRT, aiming to reduce distant relapse and enhance long-term survival. By October 2025, 60 patients with early-stage non-gastric MALT lymphoma were enrolled. Among the per-protocol efficacy-evaluable cohort (n = 55), the combined immunoradiotherapy regimen achieved a complete response rate of 100%. At a median follow-up of 30.2 months, only one distant recurrence was observed (estimated 5-year distant recurrence: 1.9%). In the full analysis set, 2- and 4-year progression-free survival rates were 98.0% and 92.9%, respectively, improving to 100% and 94.4% in the per-protocol set. Treatment-related hematologic toxicity was frequent but manageable. Infections were reported in 23.3%, including one grade 4 respiratory infection that necessitated treatment discontinuation. Immunophenotypic profiling revealed that parotid, thyroid, or mediastinal involvement correlated with higher lymphocyte proportions (39.5% ± 16.4%, P = 0.012). Post-treatment immunologic changes featured near-complete B-cell depletion and a compensatory expansion of NK cells. Overall, the combined immunoradiotherapy regimen demonstrated durable disease control in localized non-gastric MALT lymphoma, with potential synergistic benefit from NK cell-mediated immune activation. Trial registration: Chinese Clinical Trials Registry, ChiCTR2000036318; registered on Aug 22, 2020 (prospective).

Relapsed/refractory classical Hodgkin lymphoma (R/R cHL) remains clinically challenging due to substantial heterogeneity in relapse risk. Reliable, non-invasive tools for improved relapse risk stratification are urgently needed to guide individualized therapeutic strategies. In this multicenter retrospective study, 161 patients with R/R cHL from five institutions were included (training cohort: n = 102; validation cohort: n = 59). Clinical and metabolic covariates were assessed at the time of relapsed/refractory disease and baseline 18F-FDG PET/CT before salvage treatment. We developed a deep learning-based Mixture-of-Experts (MoE) framework that integrates four medical foundation models (PET-Diffusion, SAM-Med2D, MedCLIP, and RadFM) to derive a quantitative imaging biomarker (MoEScore) from baseline 18F-FDG PET/CT. A multiparametric model incorporating MoEScore and independent clinical/metabolic predictors was constructed and evaluated using discrimination, calibration, and clinical utility analyses. Model interpretability was assessed using attention visualization, ablation analysis, and pathological correlation. MoEScore demonstrated predictive performance (AUC: 0.861 in training; 0.783 in validation) and remained independently associated with relapse (HR = 11.18, 95

PurposeThis study assessed the utility of baseline 18F-FDG PET/CT radiomics by integrating tumor habitat analysis with both intra- and peritumoral features to predict EGFR mutation status in lung adenocarcinoma.MethodsA total of 724 patients from two centers were allocated to training, validation, and test cohorts. Peritumoral regions were delineated with 2–8 mm radial expansions using LIFEx, while tumor habitat subregions were identified via k-means clustering. Multiple machine learning algorithms were employed to develop clinical-metabolic, intratumoral, peritumoral, habitat, and combined models. Model performance was assessed using AUC, calibration curves, DCA, and DeLong tests, and SHAP analysis was applied to interpret critical predictive features.ResultsIn the test cohort, the combined model achieved the highest and most favorable predictive performance for EGFR mutation (AUC = 0.862, 95% CI: 0.80–0.93), followed by the habitat model (AUC = 0.831, 95% CI: 0.76–0.90). Both models significantly outperformed all other models across datasets (all P < 0.05). Among peritumoral models, the 6 mm expansion version demonstrated the highest AUC. SHAP analysis indicated that 16 of the 17 key features in the habitat model originated from Habitat 1 and 2 subregions, and approximately two-thirds of the top predictive features were CT-based.ConclusionBaseline 18F-FDG PET/CT radiomics provides reliable prediction of EGFR mutation. Both the habitat and combined models show comparable and strong predictive performance to guide image-informed personalized treatment, while SHAP analysis enhances interpretability for clinical implementation.

Background:Patients with follicular lymphoma (FL) who experience progression of disease within 24 months (POD24) of receiving first-line therapy had a significantly poorer prognosis than that without early progression. Due to the established prognostic relevance of positron emission tomography/computed tomography (PET/CT) parameters in FL and their clinical accessibility, we aimed to investigate the predictive role of PET/CT metabolism and dissemination parameters in POD24 for FL. Methods:The POD24 status of 155 patients who underwent PET/CT examinations at initial diagnosis was evaluated. Various baseline characteristics were collected, along with PET/CT-derived parameters, including the maximum tumor dissemination (Dmax), maximum standardized uptake (SUVmax) value, total metabolic tumor volume (TMTV), and total lesion glycolysis (TLG). A Cox proportional regression analysis was used to identify potential risk predictors of POD24. Receiver operating characteristic (ROC) curves were used to define the optimal cut-off values. Results:In our cohort, POD24 was observed in 21 (13.5%) FL patients. The univariate and multivariate Cox regression analyses revealed that elevated lactate dehydrogenase (LDH) was a significant predictor of POD24. Additionally, survival analyses based on the cut-off values showed that the risk of POD24 was significantly increased in patients with a Dmax >64.24 cm, SUVmax >11.23, TMTV >144.16 cm2, and TLG >586.79 g. Further, a Dmax >64.24 cm, a TMTV >144.16 cm2, and elevated LDH were selected for inclusion in a risk model [concordance index (C-index) =0.82], and the patients were divided into three risk groups, in which the rates of POD24 were 1.69%, 10.42%, and 35.29%, respectively (P<0.001). Our model exhibited excellent performance in terms of both the C-index and ROC curve analysis, surpassing the performance of models commonly used in the field. Conclusions:PET/CT parameters have prognostic value for POD24 in FL. The risk model, which combined PET/CT parameters with clinical indicators, could improve risk stratification and help guide therapeutic decisions.

OBJECTIVE:Prognosis evaluation in cervical cancer is crucial for treatment decisions. This study aims to develop and validate a combined model using positron emission tomography (PET)-derived intratumoral and peritumoral radiomic parameters to predict cervical cancer prognosis based on the Shapley additive explanations (SHAP) method. SUBJECTS AND METHODS:A retrospective cohort of 114 patients with cervical cancer from two institutions was used, with one institution's data designated for training and the other for testing. Semi-automatic segmentation of fluorine-18-fluorodeoxyglucose (18F-FDG) PET images was performed to delineate the primary intratumoral and peritumoral regions, defined by expanding the tumor boundary by 2mm, 4mm, 6mm, and 8mm. Radiomic features were extracted from each region. Six machine learning algorithms were employed to construct intratumoral and peritumoral radiomic models, with the optimal model selected based on performance evaluated through receiver operating characteristic (ROC) and calibration curves. Area under the curve (AUC) values were compared using the DeLong test. The SHAP method was used to identify the key features influencing prognosis. RESULTS:Among the intratumoral and peritumoral radiomic models, the Gradient Boosting Machine (GBM) algorithm showed superior performance. The 4mm peritumoral model exhibited the best performance among the four peritumoral models, with a testing AUC of 0.762 (95% CI: 0.582-0.944). The integrated model combining the intratumoral and 4mm peritumoral regions emerged as the optimal radiomic model for predicting cervical cancer prognosis, achieving the highest AUC of 0.954 (95% CI: 0.882-1.000) in the testing set. At the patient level, SHAP force plots provided valuable insights into the combined model's predictive ability for prognosis. CONCLUSION:The integrated radiomic model, particularly for the 4mm peritumoral region, was validated as the optimal approach for predicting overall survival in cervical cancer. The application of the SHAP method enhanced interpretability, allowing for the identification of key features influencing prognosis and offering transparent insights for guiding personalized treatment strategies.

The progression of disease (POD) within 24 months is well-established as an early indicator in various lymphomas. However, the aggressive nature of extranodal NK/T-cell lymphoma (ENKTL) necessitates a shorter assessment period. In this study, we examined POD metrics in 170 newly diagnosed ENKTL patients receiving asparaginase-based regimens. Patients were divided into three groups based on relapse/progression timing: aggressive group (POD6), progressive group (POD6-24), and stable group (POD > 24). The median overall survival (OS) in the POD6 group was 16.5 months, while the median OS was not reached in the remaining two groups. Survival outcomes were comparable between the populations with POD6 and POD24. Cox regression analysis identified POD6 as an independent risk factor for OS, regardless of traditional risk stratification. In conclusion, POD6 is a valuable early prognostic marker for ENKTL, serving as a long-term prognosis surrogate for clinical studies.

This study investigated the predictive ability of a transformer model utilizing intratumoral, peritumoral, and habitat features derived from pretreatment 18F-FDG PET imaging to assess overall survival (OS) in patients with cervical cancer. A retrospective analysis was performed using pretreatment PET data from 107 patients with cervical cancer across two medical institutions. The k-means unsupervised clustering algorithm categorized the tumor and its 4 mm peritumoral region into four distinct habitat subregions. Radiomic features were extracted from the intratumoral, peritumoral, and each habitat subregion to construct intratumoral, peritumoral, habitat, and combined transformer models. Model performance was evaluated using the area under the receiver operating characteristic (ROC) curve, calibration curves, and decision curve analysis. The habitat subregion 1 model demonstrated the highest performance. Among individual models, the habitat transformer model achieved the strongest results, with an external validation set area under the curve (AUC) of 0.778 (95

BACKGROUND:This study aimed to construct a radiomics-based imaging biomarker for the non-invasive identification of transformed follicular lymphoma (t-FL) using PET/CT images. METHODS:A total of 784 follicular lymphoma (FL), diffuse large B-cell lymphoma, and t-FL patients from 5 independent medical centers were included. The unsupervised EMFusion method was applied to fuse PET and CT images. Deep-based radiomic features were extracted from the fusion images using a deep learning model (ResNet18). These features, along with handcrafted radiomics, were utilized to construct a radiomic signature (R-signature) using automatic machine learning in the training and internal validation cohort. The R-signature was then tested for its predictive ability in the t-FL test cohort. Subsequently, this R-signature was combined with clinical parameters and SUVmax to develop a t-FL scoring system. RESULTS:The R-signature demonstrated high accuracy, with mean AUC values as 0.994 in the training cohort and 0.976 in the internal validation cohort. In the t-FL test cohort, the R-signature achieved an AUC of 0.749, with an accuracy of 75.2%, sensitivity of 68.0%, and specificity of 77.5%. Furthermore, the t-FL scoring system, incorporating the R-signature along with clinical parameters (age, LDH, and ECOG PS) and SUVmax, achieved an AUC of 0.820, facilitating the stratification of patients into low, medium, and high transformation risk groups. CONCLUSIONS:This study offers a promising approach for identifying t-FL non-invasively by radiomics analysis on PET/CT images. The developed t-FL scoring system provides a valuable tool for clinical decision-making, potentially improving patient management and outcomes.

OBJECTIVES:To develop and evaluate the predictive efficacy of a combined model incorporating clinical parameters and PET-based radiomics signature (R-signature) for prognosis in patients with metastatic melanoma. METHODS:A total of 187 metastatic melanoma patients from two centers were included, with the datasets from each center divided into training and validation cohorts, respectively. The optimal machine learning algorithm selected from the six candidates was used to construct the model. Five-fold cross-validation was performed on the training cohort for internal validation, while the external validation cohort was used for independent validation. The area under the receiver operating characteristic curve (AUC) was used to compare the model accuracies. Furthermore, multiparametric models were designed based on results from the Cox proportional hazards model and assessed through calibration curves, concordance index (C-index), and decision curve analysis (DCA) in the training and validation cohorts. RESULTS:The cutoff values for R-signature predicting progression-free survival (PFS) and overall survival (OS) were 0.47 and 0.59, respectively. The combined model showed robust prognostic performance, with C-indices of 0.92 (95%CI: 0.83-0.98) for PFS and 0.99 (95%CI: 0.97-0.99) for OS in the train cohort. Validation cohort confirmed these findings, with C-indices of 0.95 (95%CI: 0.86-0.99) for PFS and 0.97 (95%CI: 0.92-1.00) for OS. Calibration and decision curve analyses supported the clinical value of the combined model. CONCLUSION:PET-based R-signature offers valuable prognostic insight in metastatic melanoma, with the combined model further improving risk stratification. Moreover, the multiparametric models developed in this study exhibited promising potential in accurately stratifying patients based on their survival risk.

This retrospective study investigated preliminarily the prognostic value of body composition parameters derived from baseline 18F-FDG PET/CT in angioimmunoblastic T-cell lymphoma (AITL) patients. We included 94 treatment-naïve AITL patients diagnosed by histopathology. Based on the axial CT images of the third lumbar vertebra, the areas of skeletal muscle (SM), visceral adipose tissue (VAT), and subcutaneous adipose tissue (SAT) were semi-automatically delineated and standardized to finally obtain SM index (SMI), VAT index (VATI), and SAT index (SATI). Body composition density characterized by CT attenuation was obtained, including SM density (SMD), VAT density (VATD), and SAT density (SATD). Besides, the maximum standardized uptake value (SUVmax), total metabolic tumor volume (TMTV), and total lesion glycolysis (TLG) were also calculated. Endpoints included progression-free survival (PFS) and overall survival (OS). Survival curves and Cox regression analysis were performed‌‌.Of 94 AITL patients(mean age:63.4 ± 9.7 years,67

OBJECTIVE:Double and triple hit lymphoma (DHL/THL) is a rare genetic subtype of diffuse large B-cell lymphoma (DLBCL) characterized by extremely aggressive behavior. Tumor necrosis and ascites are also commonly associated with a poor prognosis. This study aimed to assess the potential prognostic markers on baseline fluorine-18-fluorodeoxyglucose (18F-FDG) positron emission tomography/computed tomography (PET/CT) in DHL/THL patients. SUBJECTS AND METHODS:A total of 36 DHL/THL patients, with myelocytomatosis (MYC) and B-cell lymphoma 2 (BCL2) and/or B-cell lymphoma 6 (BCL6) rearrangements from 3 independent medical centers, were retrospectively studied. These patients all underwent baseline 18F-FDG PET/CT scans between November 2012 and February 2024. Presence of tumor necrosis and ascites were visually assessed at PET and CT, as necrosisPET and ascitesCT, respectively. Survival analyses were performed using Cox regression and Kaplan-Meier methods. Progression-free survival (PFS) and overall survival (OS) were used as endpoints. Time-dependent receiver operating characteristics (ROC) curves were used to assess the predictive power of necrosisPET and ascitesCT. RESULTS:Of the 36 patients, 14 experienced recurrence, and 12 died during the follow-up period. Multivariate analysis revealed that necrosisPET (hazard ratio [HR]=5.272, P=0.009; HR=5.363, P=0.026) and ascitesCT (HR=5.558, P=0.008; HR=7.384, P=0.020) were independently prognostic factors for PFS and OS. Time-dependent ROC analysis showed that necrosisPET and ascitesCT had good performance in predicting PFS (accuracy: 0.671-0.758) and OS (accuracy: 0.686-0.769) occurring within 1-4 years. The risk model, incorporating necrosisPET and ascitesCT, can effectively stratify patients for PFS (χ2=11.804, P=0.003 and χ2=15.754, P<0.001, respectively) and OS (χ2=20.599, P<0.001 and χ2=18.369, P<0.001, respectively) in subgroup analyses of patients treated with R-CHOP and intensive regimens, respectively. CONCLUSION:NecrosisPET and ascitesCT independently predict survival in DHL/THL patients and may improve risk stratification, potentially guiding personalized therapeutic strategies.

RATIONALE AND OBJECTIVES:Positron emission tomography/computed tomography (PET/CT) is central to lymphoma staging, yet current guidelines still mandate bone marrow biopsy (BMB) for histologically confirming bone marrow infiltration (BMI, which defines stage IV disease) in the initial staging of extranodal NK/T-cell lymphoma (ENKTL). This study aims to compare the diagnostic accuracy and prognostic value of PET/CT and BMB in detecting BMI, and to evaluate whether BMB can be omitted in certain subgroups to refine staging protocols and avoid invasive procedures. MATERIALS AND METHODS:We conducted a retrospective cohort study of 133 treatment-naïve ENKTL patients who underwent concurrent PET/CT and BMB. RESULTS:BMB results were negative in all PET/CT-defined localized-stage patients (n=88) but positive in 15.6% (7/45) of advanced-stage cases. Using BMB as the reference standard, PET/CT showed high sensitivity (85.7%) and negative predictive value (NPV=98.9%), supporting disease exclusion. Specificity was 73.8%, accuracy 74.4%. In advanced disease, sensitivity remained 85.7%, NPV 95.7%, but specificity declined to 57.9%. Prognostically, PET/CT-defined BMI predicted inferior progression-free survival (PFS, hazard ratio [HR]=2.36, P=0.036) and overall survival (OS, HR=3.3, P=0.026), whereas BMB positivity showed limited prognostic discrimination (PFS HR=1.7, P=0.293; OS HR=3.08, P=0.039). Time-dependent area under the curve analysis demonstrated that PET/CT-detected BMI had consistent superiority in prognostic prediction across all follow-up intervals, a finding further substantiated by its higher C-index values for both PFS (0.63 vs. 0.55) and OS (0.69 vs. 0.60) compared to BMB. Notably, BMB results did not alter early/advanced staging or treatment allocation, but showed significant association with hemophagocytic lymphohistiocytosis (85.7% vs. 28.9%, P=0.015). CONCLUSION:PET/CT demonstrates high diagnostic and prognostic value in ENKTL staging, supported by an NPV of 98.9% and general alignment with therapeutic decision-making. Thus, routine BMB could potentially be omitted in the absence of unexplained cytopenias. Further validation in larger prospective cohorts is warranted before broad implementation.

Pathological grade is a critical determinant of clinical outcomes and decision-making of follicular lymphoma (FL). This study aimed to develop a deep learning model as a digital biopsy for the non-invasive identification of FL grade. This study retrospectively included 513 FL patients from five independent hospital centers, randomly divided into training, internal validation, and external validation cohorts. A multimodal fusion Transformer model was developed integrating 3D PET tumor images with tabular data to predict FL grade. Additionally, the model is equipped with explainable modules, including Gradient-weighted Class Activation Mapping (Grad-CAM) for PET images, SHapley Additive exPlanations analysis for tabular data, and the calculation of predictive contribution ratios for both modalities, to enhance clinical interpretability and reliability. The predictive performance was evaluated using the area under the receiver operating characteristic curve (AUC) and accuracy, and its prognostic value was also assessed. The Transformer model demonstrated high accuracy in grading FL, with AUCs of 0.964–0.985 and accuracies of 90.2-96.7

OBJECTIVE:Lymph node (LN) staging in lung cancer is crucial for treatment decisions. To develop and validate a positron emission tomography/computed tomography (PET/CT) radiomics model for preoperative estimation of LN metastasis in non-small cell lung cancer (NSCLC). SUBJECTS AND METHODS:A retrospective analysis of 252 NSCLC patients with 548 pathologically confirmed LN, including 227 occult LN, was performed. Clinical and PET/CT features were collected. Eight machine learning models were used for feature selection and radiomics signature (R-signature) construction. Models were developed for both the overall and occult LN groups. Model performance was evaluated using area under the curve (AUC), calibration, and decision curve analysis. RESULTS:The random forest-enhanced logistic regression (RFELR) model, based on 20 features, showed the best performance in predicting LN metastasis in both groups. The combined model demonstrated the highest predictive efficacy, with AUC of 0.94 (overall LN) and 0.89 (occult LN) in the training cohort, and 0.95 (overall LN) and 0.78 (occult LN) in the validation cohort. The combined model outperformed clinical, CT, and PET models (P<0.05) in both cohorts. Decision curve analysis showed a greater net benefit across a wider range of threshold probabilities for LN metastasis prediction. CONCLUSION:The combined model, integrating clinical, conventional PET/CT, and radiomics features, significantly enhances LN metastasis diagnosis. It shows promise in predicting occult LN metastasis and offers valuable support for personalized therapeutic decisions in NSCLC patients.