Shu Zhou
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Shu Zhou has a total of 44 co-authors. Here are the top co-authors by number of shared publications:
Yinjiao Fei
Department of Oncology, Sir Run Run Hospital, Nanjing Medical University, Nanjing, China
15 shared publicationsJinling Yuan
The First Affiliated Hospital of Nanjing Medical University
12 shared publicationsLei Qiu
Department of Oncology, Beijing Tiantan Hospital Capital Medical University Beijing China
12 shared publicationsMengxing Wu
The First Affiliated Hospital of Nanjing Medical University
12 shared publications- Yuandong Cao
Department of Radiation Oncology, The First Affiliated Hospital with Nanjing Medical University, Nanjing, China.
11 shared publications Yuchen Zhu
Department of Radiation Oncology, The First Affiliated Hospital With Nanjing Medical University, Nanjing, China
11 shared publications
Latest Publications
- Open Access
- 13 May 2026
- Published by Springer Nature in npj Precision Oncology
- Vol. 10 (1)
Radiomics model integrating MRI and ECV enhances prediction accuracy for progression in high-grade glioma
Glioma is the most common primary brain tumor, with high-grade glioma (HGG) posing significant clinical challenges due to its poor survival outcomes. One-year tumor recurrence indicates a poor prognosis, making accurate progression risk prediction models critical for clinical decision-making. This study aimed to develop a novel combined model (DL_com) based on the MobileNet-based Hybrid Network (MobHy-Net), integrating clinical variables and deep learning features from both T2-FLAIR and extracellular volume images to predict 1-year progression risk. Preoperative multi-sequence MRI (T1WI, T1C, and T2-FLAIR) from 193 HGG patients across two centers was analyzed. DL_com demonstrated superior predictive performance, with area under the curve values of 0.954 (training), 0.911 (validation), and 0.919 (test), significantly outperforming other models (P < 0.05). Furthermore, decision curve analysis confirmed its clinical utility, and Shapley Additive Explanations analysis enhanced its visualization and interpretability. DL_com effectively predicts 1-year progression risk in HGG, offering a valuable tool for risk stratification and clinical decision support.
- 13 May 2026
- Published by Springer Nature in npj Precision Oncology
- Vol. 10 (1)
Open AccessRadiomics model integrating MRI and ECV enhances prediction accuracy for progression in high-grade glioma
Glioma is the most common primary brain tumor, with high-grade glioma (HGG) posing significant clinical challenges due to its poor survival outcomes. One-year tumor recurrence indicates a poor prognosis, making accurate progression risk prediction models critical for clinical decision-making. This study aimed to develop a novel combined model (DL_com) based on the MobileNet-based Hybrid Network (MobHy-Net), integrating clinical variables and deep learning features from both T2-FLAIR and extracellular volume images to predict 1-year progression risk. Preoperative multi-sequence MRI (T1WI, T1C, and T2-FLAIR) from 193 HGG patients across two centers was analyzed. DL_com demonstrated superior predictive performance, with area under the curve values of 0.954 (training), 0.911 (validation), and 0.919 (test), significantly outperforming other models (P < 0.05). Furthermore, decision curve analysis confirmed its clinical utility, and Shapley Additive Explanations analysis enhanced its visualization and interpretability. DL_com effectively predicts 1-year progression risk in HGG, offering a valuable tool for risk stratification and clinical decision support.
- Open Access
- 6 January 2026
- Published by SAGE Publications in Digital Health
- Vol. 12
A feasibility study on predicting disease progression in high-grade gliomas using magnetic resonance imaging habitat radiomics based on response assessment in neuro-oncology (RANO) criteria
Objective: Investigating progression risk insights of high-grade gliomas through habitat radiomics analysis. Methods: A cohort of 89 patients with high-grade gliomas was enrolled, with 63 patients in the train cohort and 26 patients in the test cohort. The methodology involved delineating the region of interest (ROI) within the tumor area on magnetic resonance imaging images, followed by multisequence registration. The ROI was further divided into subregions using K -means clustering. Radiomics features were extracted from the subregions, and feature selection was performed using least absolute shrinkage and selection operator regression. Four separate models were created: radiomics model, clinical model, habitat model, and combined model that merged the habitat signature with clinical factors. The accuracy of the models was evaluated using concordance index (C-index) and receiver operating characteristic analysis. Results: The ROI was divided into three subregions, from which 36 features were extracted and selected. The habitat model, radiomics model, clinical model, and combined model were constructed by combining the extracted features with clinical data. The habitat model showed excellent predictive performance with the C-index values of 0.879 in the train cohort and 0.781 in the test cohort. Using this model, patients were classified into high-risk and low-risk groups, resulting in significantly different median progression-free survival (mPFS) times of 7 and 31 months, respectively ( P < 0.001). Stratifying the patient cohort based on isocitrate dehydrogenase status also revealed distinct mPFS outcomes for high-risk and low-risk patients. Conclusion: The habitat model demonstrated outstanding predictive performance for forecasting the progression risk of patients with high-grade gliomas.
- 6 January 2026
- Published by SAGE Publications in Digital Health
- Vol. 12
Open AccessA feasibility study on predicting disease progression in high-grade gliomas using magnetic resonance imaging habitat radiomics based on response assessment in neuro-oncology (RANO) criteria
Objective: Investigating progression risk insights of high-grade gliomas through habitat radiomics analysis. Methods: A cohort of 89 patients with high-grade gliomas was enrolled, with 63 patients in the train cohort and 26 patients in the test cohort. The methodology involved delineating the region of interest (ROI) within the tumor area on magnetic resonance imaging images, followed by multisequence registration. The ROI was further divided into subregions using K -means clustering. Radiomics features were extracted from the subregions, and feature selection was performed using least absolute shrinkage and selection operator regression. Four separate models were created: radiomics model, clinical model, habitat model, and combined model that merged the habitat signature with clinical factors. The accuracy of the models was evaluated using concordance index (C-index) and receiver operating characteristic analysis. Results: The ROI was divided into three subregions, from which 36 features were extracted and selected. The habitat model, radiomics model, clinical model, and combined model were constructed by combining the extracted features with clinical data. The habitat model showed excellent predictive performance with the C-index values of 0.879 in the train cohort and 0.781 in the test cohort. Using this model, patients were classified into high-risk and low-risk groups, resulting in significantly different median progression-free survival (mPFS) times of 7 and 31 months, respectively ( P < 0.001). Stratifying the patient cohort based on isocitrate dehydrogenase status also revealed distinct mPFS outcomes for high-risk and low-risk patients. Conclusion: The habitat model demonstrated outstanding predictive performance for forecasting the progression risk of patients with high-grade gliomas.
- Open Access
- 6 January 2026
- Published by Wiley in Annals of Clinical and Translational Neurology
- Vol. 13 (6) , 1160-1178
Unraveling the Molecular Mechanisms of Glioma Recurrence: A Study Integrating Single‐Cell and Spatial Transcriptomics
Objective: Glioma recurrence severely impacts patient prognosis, with current treatments showing limited efficacy. Traditional methods struggle to analyze recurrence mechanisms due to challenges in assessing tumor heterogeneity, spatial dynamics, and gene networks. Single‐cell combined spatial transcriptomics (ST) offers innovative solutions. Methods: We analyzed glioma mRNA data from TCGA and single‐cell and ST data from GEO. Following quality control, dimensionality reduction, clustering, and cell annotation of single‐cell sequencing data, we identified cell types exhibiting significantly aberrant distributions between primary and recurrent samples by analyzing the deviation degree of Ro/e values. Fibroblasts demonstrating the greatest intergroup differences were subsequently selected as the key cellular population for further investigation. Key differentially expressed genes (DEGs) were identified via random survival forest analysis. Drug sensitivity was assessed using GDSC. Deconvolution algorithms mapped cellular spatial distribution, while PROGENy quantified pathway activity. MISTy modeling revealed cell–cell interactions. Results: Fibroblasts were the primary recurrence‐associated subpopulation, with marker genes enriched in extracellular matrix and adhesion pathways. AEBP1, ZNF708, and TSHZ2 were identified as key genes: AEBP1/TSHZ2 correlated with poor prognosis, while ZNF708 showed an inverse trend. These genes were linked to chemosensitivity (Irinotecan, Carmustine, Vincristine, and Cisplatin). Recurrent tumors exhibited increased plasma cell infiltration, with key genes regulating IL‐17, Notch, and Toll‐like receptor pathways. Spatial analysis highlighted oligodendrocyte‐astrocyte interactions in the tumor microenvironment. Interpretation: Fibroblasts drive glioma recurrence, with AEBP1, ZNF708, and TSHZ2 predicting recurrence and chemoresistance. These genes promote immune suppression (via plasma cells) and activate recurrence pathways. Oligodendrocyte‐astrocyte interactions shape the recurrent microenvironment, suggesting new therapeutic targets.
- 6 January 2026
- Published by Wiley in Annals of Clinical and Translational Neurology
- Vol. 13 (6) , 1160-1178
Open AccessUnraveling the Molecular Mechanisms of Glioma Recurrence: A Study Integrating Single‐Cell and Spatial Transcriptomics
Objective: Glioma recurrence severely impacts patient prognosis, with current treatments showing limited efficacy. Traditional methods struggle to analyze recurrence mechanisms due to challenges in assessing tumor heterogeneity, spatial dynamics, and gene networks. Single‐cell combined spatial transcriptomics (ST) offers innovative solutions. Methods: We analyzed glioma mRNA data from TCGA and single‐cell and ST data from GEO. Following quality control, dimensionality reduction, clustering, and cell annotation of single‐cell sequencing data, we identified cell types exhibiting significantly aberrant distributions between primary and recurrent samples by analyzing the deviation degree of Ro/e values. Fibroblasts demonstrating the greatest intergroup differences were subsequently selected as the key cellular population for further investigation. Key differentially expressed genes (DEGs) were identified via random survival forest analysis. Drug sensitivity was assessed using GDSC. Deconvolution algorithms mapped cellular spatial distribution, while PROGENy quantified pathway activity. MISTy modeling revealed cell–cell interactions. Results: Fibroblasts were the primary recurrence‐associated subpopulation, with marker genes enriched in extracellular matrix and adhesion pathways. AEBP1, ZNF708, and TSHZ2 were identified as key genes: AEBP1/TSHZ2 correlated with poor prognosis, while ZNF708 showed an inverse trend. These genes were linked to chemosensitivity (Irinotecan, Carmustine, Vincristine, and Cisplatin). Recurrent tumors exhibited increased plasma cell infiltration, with key genes regulating IL‐17, Notch, and Toll‐like receptor pathways. Spatial analysis highlighted oligodendrocyte‐astrocyte interactions in the tumor microenvironment. Interpretation: Fibroblasts drive glioma recurrence, with AEBP1, ZNF708, and TSHZ2 predicting recurrence and chemoresistance. These genes promote immune suppression (via plasma cells) and activate recurrence pathways. Oligodendrocyte‐astrocyte interactions shape the recurrent microenvironment, suggesting new therapeutic targets.
- Open Access
- 4 January 2026
- Published by Wiley in Annals of Clinical and Translational Neurology
- Vol. 13 (6) , 1141-1153
Development of a Prediction Model for Progression Risk in High‐Grade Gliomas Based on Habitat Radiomics and Pathomics
Objective: To investigate the value of constructing models based on habitat radiomics and pathomics for predicting the risk of progression in high‐grade gliomas. Methods: This study conducted a retrospective analysis of preoperative magnetic resonance (MR) images and pathological sections from 72 patients diagnosed with high‐grade gliomas (52 cases as a train cohort and 20 cases as a test cohort). The regions of interest (ROIs) were annotated accordingly. In MRI processing, the ROI was further divided into clusters to extract habitat radiomics features. For whole slide imaging (WSI), the ROI was cropped into equal‐sized image patches for weakly supervised learning and deep learning using various network architectures. The optimal model architecture was selected, and pathological features were extracted. After feature selection, four independent models were constructed: habitat radiomics model, pathomics‐based model, clinical model, and combined model integrating all information. Model performance was evaluated using the concordance index (C‐index) and the area under the receiver operating characteristic curve (AUC). Results: The combined model demonstrated the best predictive performance, with a C‐index of 0.883 and an AUC of 0.965 in the train cohort. In the test cohort, the C‐index was 0.840, and the AUC was 0.927. Based on the combined model, patients with high‐grade gliomas were divided into high‐risk and low‐risk groups, with median progression‐free survival (mPFS) of 9 months and 77 months, respectively ( p < 0.001). Conclusion: Compared with the habitat radiomics model or the pathomics‐based model alone, the combined model can better predict the risk of progression in high‐grade gliomas and provides valuable guidance for personalized treatment of high‐grade gliomas.
- 4 January 2026
- Published by Wiley in Annals of Clinical and Translational Neurology
- Vol. 13 (6) , 1141-1153
Open AccessDevelopment of a Prediction Model for Progression Risk in High‐Grade Gliomas Based on Habitat Radiomics and Pathomics
Objective: To investigate the value of constructing models based on habitat radiomics and pathomics for predicting the risk of progression in high‐grade gliomas. Methods: This study conducted a retrospective analysis of preoperative magnetic resonance (MR) images and pathological sections from 72 patients diagnosed with high‐grade gliomas (52 cases as a train cohort and 20 cases as a test cohort). The regions of interest (ROIs) were annotated accordingly. In MRI processing, the ROI was further divided into clusters to extract habitat radiomics features. For whole slide imaging (WSI), the ROI was cropped into equal‐sized image patches for weakly supervised learning and deep learning using various network architectures. The optimal model architecture was selected, and pathological features were extracted. After feature selection, four independent models were constructed: habitat radiomics model, pathomics‐based model, clinical model, and combined model integrating all information. Model performance was evaluated using the concordance index (C‐index) and the area under the receiver operating characteristic curve (AUC). Results: The combined model demonstrated the best predictive performance, with a C‐index of 0.883 and an AUC of 0.965 in the train cohort. In the test cohort, the C‐index was 0.840, and the AUC was 0.927. Based on the combined model, patients with high‐grade gliomas were divided into high‐risk and low‐risk groups, with median progression‐free survival (mPFS) of 9 months and 77 months, respectively ( p < 0.001). Conclusion: Compared with the habitat radiomics model or the pathomics‐based model alone, the combined model can better predict the risk of progression in high‐grade gliomas and provides valuable guidance for personalized treatment of high‐grade gliomas.
- Open Access
- 2 December 2025
- Published by Springer Nature in Insights into Imaging
- Vol. 16 (1) , 267
Imaging-based transformer model predicts early therapy response in advanced nasopharyngeal carcinoma: a dual-center study
Introduction: Deep-learning methodologies for predicting early response in locally advanced nasopharyngeal carcinoma (LA-NPC) remain unvalidated, with techniques like 2.5-D imaging and Transformers underexplored. Materials and methods: MRI images from LA-NPC patients diagnosed between January 2020 and March 2024 at two centers were analyzed. Patients ( n = 184) were split into training ( n = 89), validation ( n = 39), and test ( n = 56) sets. Three segmentation models—SegResNet, Unet, and UnetR—automatically delineated regions of interest (ROIs). A 2.5D approach integrated adjacent tumor sections into a transfer learning framework, leading to three predictive models: Clinical, Transformer, and Combined. Performance was assessed using ROC curves, calibration curves, and decision curve analysis (DCA). Results: The Transformer model outperformed others, achieving AUCs of 0.968, 0.957, and 0.830 for the training, validation, and test sets, respectively. The Clinical model had lower AUCs (0.898, 0.759, 0.658). The Combined model, integrating clinical data, matched or exceeded Transformer performance, particularly in the test set (AUC = 0.874). Conclusion: The Combined model, leveraging Transformer architecture and clinical factors, demonstrates strong efficacy in predicting early response in LA-NPC patients undergoing chemoradiotherapy, suggesting its potential for improved personalized treatment. Critical relevance statement: This study critically validates a novel 2.5-D radiomic-Transformer fusion model that improves early response prediction for locally advanced nasopharyngeal carcinoma, directly advancing personalized chemoradiotherapy planning in clinical radiology. Key Points: Early treatment response prediction in locally advanced nasopharyngeal carcinoma lacks validated deep learning models using 2.5D imaging and Transformers. Transformer-based model achieved superior predictive performance compared to clinical or combined models. Integrating clinical data with Transformer imaging analysis improves personalized chemoradiotherapy outcome prediction. Graphical
- 2 December 2025
- Published by Springer Nature in Insights into Imaging
- Vol. 16 (1) , 267
Open AccessImaging-based transformer model predicts early therapy response in advanced nasopharyngeal carcinoma: a dual-center study
Introduction: Deep-learning methodologies for predicting early response in locally advanced nasopharyngeal carcinoma (LA-NPC) remain unvalidated, with techniques like 2.5-D imaging and Transformers underexplored. Materials and methods: MRI images from LA-NPC patients diagnosed between January 2020 and March 2024 at two centers were analyzed. Patients ( n = 184) were split into training ( n = 89), validation ( n = 39), and test ( n = 56) sets. Three segmentation models—SegResNet, Unet, and UnetR—automatically delineated regions of interest (ROIs). A 2.5D approach integrated adjacent tumor sections into a transfer learning framework, leading to three predictive models: Clinical, Transformer, and Combined. Performance was assessed using ROC curves, calibration curves, and decision curve analysis (DCA). Results: The Transformer model outperformed others, achieving AUCs of 0.968, 0.957, and 0.830 for the training, validation, and test sets, respectively. The Clinical model had lower AUCs (0.898, 0.759, 0.658). The Combined model, integrating clinical data, matched or exceeded Transformer performance, particularly in the test set (AUC = 0.874). Conclusion: The Combined model, leveraging Transformer architecture and clinical factors, demonstrates strong efficacy in predicting early response in LA-NPC patients undergoing chemoradiotherapy, suggesting its potential for improved personalized treatment. Critical relevance statement: This study critically validates a novel 2.5-D radiomic-Transformer fusion model that improves early response prediction for locally advanced nasopharyngeal carcinoma, directly advancing personalized chemoradiotherapy planning in clinical radiology. Key Points: Early treatment response prediction in locally advanced nasopharyngeal carcinoma lacks validated deep learning models using 2.5D imaging and Transformers. Transformer-based model achieved superior predictive performance compared to clinical or combined models. Integrating clinical data with Transformer imaging analysis improves personalized chemoradiotherapy outcome prediction. Graphical
- Open Access
- 1 October 2025
- Published by AME Publishing Company in Translational Cancer Research
- Vol. 14 (10) , 7037-7052
Development and validation of a machine learning model for predicting 3-year overall survival in metastatic nasopharyngeal carcinoma: a SEER database and web visualization study
Background: Current prognostic models for metastatic nasopharyngeal carcinoma (M1-NPC) often employ oversimplified “catch-all” classifications that fail to account for the substantial heterogeneity in metastatic patterns and treatment responses. Furthermore, existing tools lack interactive visualization capabilities to support clinical decision-making. In this study, we aim to develop and validate a visual prognostic model for 3-year overall survival (OS) in M1-NPC patients. Methods: We retrospectively analyzed clinical and pathological data from M1-NPC patients in the Surveillance, Epidemiology, and End Results (SEER) database diagnosed between 2010 and 2021 with complete follow-up. Exclusion criteria included missing surgical resection data or undocumented metastatic sites. Patients were randomly allocated to training (70%) and testing (30%) cohorts. Using univariate Cox regression, we identified significant prognostic variables among 19 clinical factors. Five machine learning (ML) algorithms—support vector machine (SVM), random forest (RF), extreme gradient boosting (XGBoost), gradient boosting decision tree (GBDT), and k-nearest neighbor (KNN)—were developed and evaluated using the area under the receiver operating characteristic (ROC) curve (AUC), sensitivity, specificity, and accuracy metrics. The top 2 algorithms were combined to form an ensemble model. All six models were tested, with SHapley Additive exPlanations (SHAP) analysis applied for interpretability. The optimal model was implemented in an online calculator for 3-year OS prediction. Results: Among 19 candidate variables, univariate Cox regression identified 11 significant prognostic factors (PConclusions: Our validated RF-based model addresses a critical gap in M1-NPC prognostication, offering clinicians an interpretable tool for survival prediction. While limited by database constraints, this represents the first SEER-derived online calculator for M1-NPC with immediate clinical applicability.
- 1 October 2025
- Published by AME Publishing Company in Translational Cancer Research
- Vol. 14 (10) , 7037-7052
Open AccessDevelopment and validation of a machine learning model for predicting 3-year overall survival in metastatic nasopharyngeal carcinoma: a SEER database and web visualization study
Background: Current prognostic models for metastatic nasopharyngeal carcinoma (M1-NPC) often employ oversimplified “catch-all” classifications that fail to account for the substantial heterogeneity in metastatic patterns and treatment responses. Furthermore, existing tools lack interactive visualization capabilities to support clinical decision-making. In this study, we aim to develop and validate a visual prognostic model for 3-year overall survival (OS) in M1-NPC patients. Methods: We retrospectively analyzed clinical and pathological data from M1-NPC patients in the Surveillance, Epidemiology, and End Results (SEER) database diagnosed between 2010 and 2021 with complete follow-up. Exclusion criteria included missing surgical resection data or undocumented metastatic sites. Patients were randomly allocated to training (70%) and testing (30%) cohorts. Using univariate Cox regression, we identified significant prognostic variables among 19 clinical factors. Five machine learning (ML) algorithms—support vector machine (SVM), random forest (RF), extreme gradient boosting (XGBoost), gradient boosting decision tree (GBDT), and k-nearest neighbor (KNN)—were developed and evaluated using the area under the receiver operating characteristic (ROC) curve (AUC), sensitivity, specificity, and accuracy metrics. The top 2 algorithms were combined to form an ensemble model. All six models were tested, with SHapley Additive exPlanations (SHAP) analysis applied for interpretability. The optimal model was implemented in an online calculator for 3-year OS prediction. Results: Among 19 candidate variables, univariate Cox regression identified 11 significant prognostic factors (PConclusions: Our validated RF-based model addresses a critical gap in M1-NPC prognostication, offering clinicians an interpretable tool for survival prediction. While limited by database constraints, this represents the first SEER-derived online calculator for M1-NPC with immediate clinical applicability.
- Open Access
- 1 October 2025
- Published by AME Publishing Company in Translational Cancer Research
- Vol. 14 (10) , 7102-7118
Efficacy, safety and cost-effectiveness of chemo-immunotherapy combinations for recurrent or metastatic nasopharyngeal carcinoma: an updated systematic review and cost-effectiveness analysis
Background: While toripalimab, camrelizumab and tislelizumab plus gemcitabine and cisplatin (TorGP, CamGP and TisGP) have been approved as first-line therapy for recurrent or metastatic nasopharyngeal carcinoma (R/M NPC) in China, comprehensive comparisons of their relative clinical and economic value remain unavailable. Therefore, the objective of this study is to identify the most beneficial first-line treatment strategy for R/M NPC that balances efficacy, safety, and cost-effectiveness, thereby supporting informed clinical decision-making and efficient healthcare resource allocation. Methods: A comprehensive systematic search was conducted across multiple databases—including PubMed, Embase, the Cochrane Library, and ClinicalTrials.gov—for phase 3 randomized controlled trials (RCTs) evaluating chemo-immunotherapy combinations in R/M NPC, covering all records from database inception through June 20, 2025. The study protocol has been formally registered with PROSPERO under the registration number CRD42023422137. We conduct a network meta-analysis (NMA), with the primary outcomes being hazard ratios (HRs) for overall survival (OS) and progression-free survival (PFS). For safety assessment, we evaluate the incidence of treatment-emergent adverse events (TEAEs). For the cost-effectiveness analysis (CEA), we develop a partitioned survival model with a 10-year time horizon and three health states: PFS, progressed disease, and death, using 3-week cycles. As a scenario analysis, we incorporate restricted cubic splines (RCS) and mixture cure model (MCM). The primary outcomes of the CEA includ quality-adjusted life-years (QALYs), life-years (LYs), and incremental cost-effectiveness ratios (ICERs). We perform extensive sensitivity analyses to verify the robustness of our findings. Methodological rigor of incorporated RCTs was assessed utilizing the Cochrane Risk of Bias Tool (RoB 2.0). Results: Three RCTs—JUPITER-02, CAPTAIN-1st, and RATIONALE-309—collectively encompassing 815 patients, formed the basis of the NMA. Based on the surface under the cumulative ranking curve estimated relative ranking in the NMA, TisGP may have higher efficacy and safety compared with other treatments, but there is no statistical significance. In terms of CEA, based on standard parametric survival model, we add RCS and MCM as scenario analysis and yielded consistent conclusions. In the standard parametric survival model, CamGP, TorGP, and TisGP were associated with ICERs of $24,331/QALY, $18,776/QALY, and $20,762/QALY, respectively, compared to GP. Compared to CamGP, TisGP and TorGP had ICERs of $5,880/QALY and $10,567/QALY, respectively. Compared to TorGP, TisGP has ICERs of $67,564/QALY, which is higher than the willingness-to-pay (WTP) threshold of $40,354.20/QALY. As a result, TorGP is proven to be the most cost-effective option. TorGP to be cost-effective under a WTP of 3-time gross domestic product (GDP) per capita is 100%. Conclusions: In summary, while TisGP shows a clinically favorable trend in efficacy and safety, TorGP demonstrates superior cost-effectiveness with an incremental cost-effectiveness ratio well below the willingness-to-pay threshold. This finding suggests that clinicians should comprehensively evaluate individual patient characteristics when selecting treatment regimens.
- 1 October 2025
- Published by AME Publishing Company in Translational Cancer Research
- Vol. 14 (10) , 7102-7118
Open AccessEfficacy, safety and cost-effectiveness of chemo-immunotherapy combinations for recurrent or metastatic nasopharyngeal carcinoma: an updated systematic review and cost-effectiveness analysis
Background: While toripalimab, camrelizumab and tislelizumab plus gemcitabine and cisplatin (TorGP, CamGP and TisGP) have been approved as first-line therapy for recurrent or metastatic nasopharyngeal carcinoma (R/M NPC) in China, comprehensive comparisons of their relative clinical and economic value remain unavailable. Therefore, the objective of this study is to identify the most beneficial first-line treatment strategy for R/M NPC that balances efficacy, safety, and cost-effectiveness, thereby supporting informed clinical decision-making and efficient healthcare resource allocation. Methods: A comprehensive systematic search was conducted across multiple databases—including PubMed, Embase, the Cochrane Library, and ClinicalTrials.gov—for phase 3 randomized controlled trials (RCTs) evaluating chemo-immunotherapy combinations in R/M NPC, covering all records from database inception through June 20, 2025. The study protocol has been formally registered with PROSPERO under the registration number CRD42023422137. We conduct a network meta-analysis (NMA), with the primary outcomes being hazard ratios (HRs) for overall survival (OS) and progression-free survival (PFS). For safety assessment, we evaluate the incidence of treatment-emergent adverse events (TEAEs). For the cost-effectiveness analysis (CEA), we develop a partitioned survival model with a 10-year time horizon and three health states: PFS, progressed disease, and death, using 3-week cycles. As a scenario analysis, we incorporate restricted cubic splines (RCS) and mixture cure model (MCM). The primary outcomes of the CEA includ quality-adjusted life-years (QALYs), life-years (LYs), and incremental cost-effectiveness ratios (ICERs). We perform extensive sensitivity analyses to verify the robustness of our findings. Methodological rigor of incorporated RCTs was assessed utilizing the Cochrane Risk of Bias Tool (RoB 2.0). Results: Three RCTs—JUPITER-02, CAPTAIN-1st, and RATIONALE-309—collectively encompassing 815 patients, formed the basis of the NMA. Based on the surface under the cumulative ranking curve estimated relative ranking in the NMA, TisGP may have higher efficacy and safety compared with other treatments, but there is no statistical significance. In terms of CEA, based on standard parametric survival model, we add RCS and MCM as scenario analysis and yielded consistent conclusions. In the standard parametric survival model, CamGP, TorGP, and TisGP were associated with ICERs of $24,331/QALY, $18,776/QALY, and $20,762/QALY, respectively, compared to GP. Compared to CamGP, TisGP and TorGP had ICERs of $5,880/QALY and $10,567/QALY, respectively. Compared to TorGP, TisGP has ICERs of $67,564/QALY, which is higher than the willingness-to-pay (WTP) threshold of $40,354.20/QALY. As a result, TorGP is proven to be the most cost-effective option. TorGP to be cost-effective under a WTP of 3-time gross domestic product (GDP) per capita is 100%. Conclusions: In summary, while TisGP shows a clinically favorable trend in efficacy and safety, TorGP demonstrates superior cost-effectiveness with an incremental cost-effectiveness ratio well below the willingness-to-pay threshold. This finding suggests that clinicians should comprehensively evaluate individual patient characteristics when selecting treatment regimens.
- Open Access
- 11 August 2025
- Published by Frontiers Media SA in Frontiers in Neurology
- Vol. 16, 1614678
Deep learning and pathomics analyses predict prognosis of high-grade gliomas
Objective: Utilizing pathomics to analyze high-grade gliomas and provide prognostic insights.Methods: Regions of Interest (ROIs) in tumor areas were identified in whole-slide images (WSI). Tumor patches underwent cropping, white space removal, and normalization. A deep learning model trained on these patches aggregated predictions for WSIs. Pathological features were extracted using Pearson correlation, univariate Cox regression, and LASSO-Cox regression. Three models were developed: a Pathomics-based model, a clinical model, and a combined model integrating both.Results: Pathological and Clinical Features were used to build two models, leading to a predictive model with a C-index of 0.847 (train) and 0.739 (test). High-risk patients had a median progression-free survival (PFS) of 10 months (p<0.001), while low-risk patients had not reached median PFS. Stratification by IDH status revealed significant PFS differences.Conclusion: The combined model effectively predicts high-grade glioma prognosis.
- 11 August 2025
- Published by Frontiers Media SA in Frontiers in Neurology
- Vol. 16, 1614678
Open AccessDeep learning and pathomics analyses predict prognosis of high-grade gliomas
Objective: Utilizing pathomics to analyze high-grade gliomas and provide prognostic insights.Methods: Regions of Interest (ROIs) in tumor areas were identified in whole-slide images (WSI). Tumor patches underwent cropping, white space removal, and normalization. A deep learning model trained on these patches aggregated predictions for WSIs. Pathological features were extracted using Pearson correlation, univariate Cox regression, and LASSO-Cox regression. Three models were developed: a Pathomics-based model, a clinical model, and a combined model integrating both.Results: Pathological and Clinical Features were used to build two models, leading to a predictive model with a C-index of 0.847 (train) and 0.739 (test). High-risk patients had a median progression-free survival (PFS) of 10 months (p<0.001), while low-risk patients had not reached median PFS. Stratification by IDH status revealed significant PFS differences.Conclusion: The combined model effectively predicts high-grade glioma prognosis.
- Open Access
- 26 May 2025
- Published by Springer Nature in Discover Oncology
- Vol. 16 (1) , 1-14
A prognostic model for head and neck squamous cell carcinoma based on radiotherapy sensitivity insights from nasopharyngeal carcinoma
Background Radioresistance significantly impairs treatment efficacy and prognostic outcomes in head and neck squamous cell carcinoma (HNSCC). This study aimed to identify radiotherapy sensitivity-related genes and construct a prognostic model for HNSCC, incorporating insights from nasopharyngeal carcinoma (NPC) as a related subtype. Methods Differentially expressed genes (DEGs) associated with radiotherapy response were identified using the GSE48501 dataset, primarily derived from NPC. Functional annotation was performed via Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses. Using the TCGA-HNSC dataset, we developed a prognostic risk model through univariate and LASSO-Cox regression analyses. The model was validated for prognostic accuracy and further analyzed for associations with immune cell infiltration, drug sensitivity, and survival outcomes using CIBERSORT, TIMER, Genomics of Drug Sensitivity in Cancer (GDSC), and nomogram analysis. Results We identified 263 DEGs related to radiotherapy sensitivity and developed a robust prognostic model based on 8 hub genes. The model effectively stratified patients into high- and low-risk groups, with superior overall survival (OS) observed in the low-risk group. The Receiver Operating Characteristic (ROC) analysis confirmed high predictive accuracy for 1-, 3-, and 5-year OS. Immune infiltration analysis revealed reduced immune activity in the high-risk group, while drug sensitivity analysis highlighted potential therapeutic strategies. The nomogram further demonstrated excellent predictive performance. Conclusion This study bridges insights from NPC-derived DEGs and HNSCC prognostic modeling, emphasizing radiotherapy sensitivity and integrating immune and therapeutic dimensions. The resulting model offers a novel approach to improve prognostic accuracy and guide treatment strategies for HNSCC patients.
- 26 May 2025
- Published by Springer Nature in Discover Oncology
- Vol. 16 (1) , 1-14
Open AccessA prognostic model for head and neck squamous cell carcinoma based on radiotherapy sensitivity insights from nasopharyngeal carcinoma
Background Radioresistance significantly impairs treatment efficacy and prognostic outcomes in head and neck squamous cell carcinoma (HNSCC). This study aimed to identify radiotherapy sensitivity-related genes and construct a prognostic model for HNSCC, incorporating insights from nasopharyngeal carcinoma (NPC) as a related subtype. Methods Differentially expressed genes (DEGs) associated with radiotherapy response were identified using the GSE48501 dataset, primarily derived from NPC. Functional annotation was performed via Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses. Using the TCGA-HNSC dataset, we developed a prognostic risk model through univariate and LASSO-Cox regression analyses. The model was validated for prognostic accuracy and further analyzed for associations with immune cell infiltration, drug sensitivity, and survival outcomes using CIBERSORT, TIMER, Genomics of Drug Sensitivity in Cancer (GDSC), and nomogram analysis. Results We identified 263 DEGs related to radiotherapy sensitivity and developed a robust prognostic model based on 8 hub genes. The model effectively stratified patients into high- and low-risk groups, with superior overall survival (OS) observed in the low-risk group. The Receiver Operating Characteristic (ROC) analysis confirmed high predictive accuracy for 1-, 3-, and 5-year OS. Immune infiltration analysis revealed reduced immune activity in the high-risk group, while drug sensitivity analysis highlighted potential therapeutic strategies. The nomogram further demonstrated excellent predictive performance. Conclusion This study bridges insights from NPC-derived DEGs and HNSCC prognostic modeling, emphasizing radiotherapy sensitivity and integrating immune and therapeutic dimensions. The resulting model offers a novel approach to improve prognostic accuracy and guide treatment strategies for HNSCC patients.
- Open Access
- 1 April 2025
- Published by SAGE Publications in Science Progress
- Vol. 108 (2)
Radiomic analysis based on machine learning of multi-sequences MR to assess early treatment response in locally advanced nasopharyngeal carcinoma
Objective: The prediction of early response in locally advanced nasopharyngeal carcinoma (LA-NPC) after concurrent chemoradiotherapy (CCRT) is important for determining the need for timely consolidation therapy. We developed a radiomic analysis of multi-sequences MR based on machine learning (ML) to assess early response in LA-NPC after CCRT. Methods: This study retrospectively enrolled 104 LA-NPC patients, randomly divided into training (70%) and test (30%) cohorts. Radiomic features were extracted from five MR sequences (T1, T1C, T2, DWI, and ADC). Feature selection was performed using Pearson's correlation coefficient and LASSO regression to reduce redundancy. ML algorithms were compared to develop models, with suboptimal sequences excluded from the multi-sequence MR fusion model. A combined model integrating the fusion and clinical model was developed using logistic regression, and its diagnostic effectiveness was evaluated using receiver operating characteristic (ROC) analysis. Results: In the mono-sequence MR analysis, T1 demonstrated the lowest discriminative capacity (AUC = 0.505), followed by T2 (AUC = 0.738). Consequently, we developed a fusion model incorporating ADC, DWI, and T1C features while excluding T1 and T2. In the test cohort, the combined model outperformed both the clinical (AUC = 0.852) and fusion (AUC = 0.886) models, achieving superior effectiveness (AUC = 0.900). Shapley Additive Explanations (SHAP) analysis identified lbp_GrayLevelVariance_ADC as the most influential predictive feature. Conclusions: A combined model, which merges clinical and multi-sequences MR radiomics model, showed good performance for predicting early response of LA-NPC after CCRT.
- 1 April 2025
- Published by SAGE Publications in Science Progress
- Vol. 108 (2)
Open AccessRadiomic analysis based on machine learning of multi-sequences MR to assess early treatment response in locally advanced nasopharyngeal carcinoma
Objective: The prediction of early response in locally advanced nasopharyngeal carcinoma (LA-NPC) after concurrent chemoradiotherapy (CCRT) is important for determining the need for timely consolidation therapy. We developed a radiomic analysis of multi-sequences MR based on machine learning (ML) to assess early response in LA-NPC after CCRT. Methods: This study retrospectively enrolled 104 LA-NPC patients, randomly divided into training (70%) and test (30%) cohorts. Radiomic features were extracted from five MR sequences (T1, T1C, T2, DWI, and ADC). Feature selection was performed using Pearson's correlation coefficient and LASSO regression to reduce redundancy. ML algorithms were compared to develop models, with suboptimal sequences excluded from the multi-sequence MR fusion model. A combined model integrating the fusion and clinical model was developed using logistic regression, and its diagnostic effectiveness was evaluated using receiver operating characteristic (ROC) analysis. Results: In the mono-sequence MR analysis, T1 demonstrated the lowest discriminative capacity (AUC = 0.505), followed by T2 (AUC = 0.738). Consequently, we developed a fusion model incorporating ADC, DWI, and T1C features while excluding T1 and T2. In the test cohort, the combined model outperformed both the clinical (AUC = 0.852) and fusion (AUC = 0.886) models, achieving superior effectiveness (AUC = 0.900). Shapley Additive Explanations (SHAP) analysis identified lbp_GrayLevelVariance_ADC as the most influential predictive feature. Conclusions: A combined model, which merges clinical and multi-sequences MR radiomics model, showed good performance for predicting early response of LA-NPC after CCRT.