The Role of Pretreatment 18F-FDG PET/CT for Early Prediction of Neoadjuvant Chemotherapy Response in Patients with Locoregionally Advanced Nasopharyngeal Carcinoma

Development of a neoadjuvant chemotherapy efficacy prediction model for nasopharyngeal carcinoma integrating magnetic resonance radiomics and pathomics: a multi-center retrospective study

OBJECTIVE

This study aimed to develop and validate a predictive model for assessing the efficacy of neoadjuvant chemotherapy (NACT) in nasopharyngeal carcinoma (NPC) by integrating radiomics and pathomics features using a particle swarm optimization-supported support vector machine (PSO-SVM).

METHODS

A retrospective multi-center study was conducted, which included 389 NPC patients who received NACT from three institutions. Radiomics features were extracted from magnetic resonance imaging scans, while pathomics features were derived from histopathological images. A total of 2,667 radiomics features and 254 pathomics features were initially extracted. Feature selection involved intra-class correlation coefficient evaluation, Mann-Whitney U test, Spearman correlation analysis, and least absolute shrinkage and selection operator regression. The PSO-SVM model was constructed and validated using 10-fold cross-validation on the training set and further evaluated using an external validation set. Model performance was assessed using the area under the curve (AUC) of the receiver operating characteristic curve, calibration curves, and decision curve analysis.

RESULTS

Eight significant predictive features (five radiomics and three pathomics) were identified. The PSO-SVM radiopathomics model achieved superior performance compared to models based solely on radiomics or pathomics features. The AUCs for the PSO-SVM radiopathomics model were 0.917 (95% CI: 0.887-0.948) in internal validation and 0.814 (95% CI: 0.742-0.887) in external validation. Calibration curves demonstrated good agreement between predicted probabilities and actual outcomes. Decision curve analysis showed that the PSO-SVM radiopathomics model provided higher clinical net benefit over a wider range of risk thresholds compared to other models.

CONCLUSION

The PSO-SVM radiopathomics model effectively integrates radiomics and pathomics features, offering enhanced predictive accuracy and clinical utility for assessing NACT efficacy in NPC. The multi-center approach and robust validation underscore its potential for personalized treatment planning, supporting improved clinical decision-making for NPC patients.

Comparative evaluation of machine learning models in predicting overall survival for nasopharyngeal carcinoma using 18F-FDG PET-CT parameters

PURPOSE

The objective of this study is to assess the prognostic efficacy of 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography/computed tomography (PET-CT) parameters in nasopharyngeal carcinoma (NPC) and identify the best machine learning (ML) prognostic model for NPC patients based on these 18F-FDG PET/CT parameters and clinical variables.

METHOD

A cohort of 678 patients diagnosed with NPC between 2016 and 2020 was analyzed in this study. The model was constructed using four advanced ML algorithms, namely Random Forest (RF), Extreme Gradient Boosting (XGBoost), Least Absolute Shrinkage and Selection Operator (LASSO), and multifactor COX step-up regression. Statistical significance of the models was assessed using Kaplan-Meier (K-M) curves, with a significance level established at P < 0.05. The prognostic efficacy of the models was evaluated through the analysis of receiver operating characteristic (ROC) curves, with the area under the ROC curve (AUC) serving as a criterion for model selection. The decision curve analysis (DCA) and concordance index (C-index) were employed to assess the precision of the optimal model.

RESULTS

Multivariate analysis revealed age, T stage, and metabolic tumor volume (MTV) for the primary nasopharyngeal tumor (MTVT) as significant independent prognostic factors for overall survival (OS) in NPC patients. Additionally, the LASSO model identified six key variables, including peak standardized uptake value (SUV-peak) for the primary nasopharyngeal tumor (SUV-peak(T)), MTVT, heterogeneity index for neck lymph nodes (HIN), age, pathological type, and T stage. Remarkably, the LASSO model demonstrated superior performance with a 5-year AUC of 0.849 compared to other models. Further assessment using the C-index and DCA confirmed the accuracy of the LASSO model. Subgroup analysis revealed notable risk factors, such as a high heterogeneity index (HI) for the primary nasopharyngeal tumor (HIT), MTV values for neck lymph nodes (MTVN), and HIN.

CONCLUSIONS

We developed a novel prognostic machine learning model that integrates 18F-FDG PET-CT parameters and clinical characteristics, significantly enhancing prognosis prediction in NPC.

A dynamic nomogram combining tumor stage and magnetic resonance imaging features to predict the response to induction chemotherapy in locally advanced nasopharyngeal carcinoma

OBJECTIVES

To establish an effective dynamic nomogram combining magnetic resonance imaging (MRI) findings of primary tumor and regional lymph nodes with tumor stage for the pretreatment prediction of induction chemotherapy (IC) response in locoregionally advanced nasopharyngeal carcinoma (LANPC).

METHODS

A total of 498 LANPC patients (372 in the training and 126 in the validation cohort) with MRI information were enrolled. All patients were classified as "favorable responders" and "unfavorable responders" according to tumor response to IC. A nomogram for IC response was built based on the results of the logistic regression model. Also, the Cox regression analysis was used to identify the independent prognostic factors of disease-free survival (DFS).

RESULTS

After two cycles of IC, 340 patients were classified as "favorable responders" and 158 patients as "unfavorable responders." Calibration curves revealed satisfactory agreement between the predicted and the observed probabilities. The nomogram achieved an AUC of 0.855 (95% CI, 0.781-0.930) for predicting IC response, which outperformed TNM staging (AUC, 0.661; 95% CI 0.565-0.758) and the MRI feature-based model alone (AUC, 0.744; 95% CI 0.650-0.839) in the validation cohort. The nomogram was used to categorize patients into high- and low-response groups. An online dynamic model was built ( https://nomogram-for-icresponse-prediction.shinyapps.io/DynNomapp/ ) to facilitate the application of the nomogram. In the Cox multivariate analysis, clinical stage, tumor necrosis, EBV DNA levels, and cervical lymph node numbers were independently associated with DFS.

CONCLUSIONS

The comprehensive nomogram incorporating MRI features and tumor stage could assist physicians in predicting IC response and formulating personalized treatment strategies for LANPC patients.

KEY POINTS

• The nomogram can predict IC response in endemic LANPC. • The nomogram combining tumor stage with MRI-based tumor features showed very good predictive performance. • The nomogram was transformed into a web-based dynamic model to optimize clinical application.