Related Experiment Video
Updated: Jun 16, 2026
A 3D Digital Model for the Diagnosis and Treatment of Pulmonary Nodules
Published on: May 19, 2023
Development of a Prognostic Predictive Model for Stage III Non-Small Cell Lung Cancer using PET-CT Radiomics Based on
Jianyang Zhang1, Yujing Hu2, Congna Tian3
1Jianyang Zhang, Hebei Medical University, Shijiazhuang 050000, Hebei, China. Department of Nuclear Medicine, Hebei General Hospital, Shijiazhuang 050000, Hebei, China.
Objective:
To compare the prognostic performance of models integrating PET/CT radiomics with metabolic parameters for predicting overall survival in patients with stage III NSCLC using multi-center retrospective data.
Methodology:
In this retrospective, comparative study, we analyzed data from 165 stage III NSCLC patients (from 37 centers in The Cancer Imaging Archive) and an independent external cohort of 48 patients from Hebei General Hospital (2022-2025). Radiomic features and metabolic parameters were extracted from baseline PET/CT images. Feature selection and cross-validation were performed to construct and compare multiple prediction models. Model performance was evaluated in training, validation, and external test cohorts using discrimination (AUC), calibration (Brier score), clinical net benefit (Decision Curve Analysis), and Kaplan-Meier survival analysis. Shapley additive explanations (SHAP) were applied to interpret the final model.
Results:
Five prediction models were developed and compared. The model combining PET radiomic features and whole-body metabolic tumor volume (MTVwb), termed the Combined_A model, demonstrated the best and most stable performance, with AUCs of 0.782, 0.775, and 0.778 in the training, validation, and external test cohorts, respectively. In the external test cohort. Combined_A also showed the lowest Brier score (Brier score: 0.188) and greater net clinical benefit. Kaplan-Meier analysis confirmed significant survival stratification (log-rank P= 0.002). SHAP analysis identified wavelet-HHL_GLSZM_Small Area Low Gray Level Emphasis and log first order Skewness as key prognostic features.
Conclusion:
In this retrospective comparative study, a model integrating PET radiomics with MTVwb provided robust and generalizable survival prediction for stage III NSCLC. These findings support its potential for personalized risk stratification, warranting further prospective validation.Registration No.: (ClinicalTrials.gov: NCT00083083).