Objective
This study aims to develop and validate a CT-based radiomics model for predicting the prognosis of head and neck cancer patients, particularly those with nasopharyngeal carcinoma (NPC), following intensity-modulated radiation therapy (IMRT).
Methods
We conducted a retrospective analysis involving 392 pathologically confirmed NPC patients from two centers. Center A contributed 226 patients to the training cohort, while Center B provided 64 patients for the validation cohort. Features extracted from CT images were utilized to develop two predictive models: a clinically combined radiomics model and a standalone radiomics model. Dimensionality reduction and nested cross-validation were employed in the model development process. The performance of the models was assessed and validated using receiver operating characteristic (ROC) curves and decision curve analysis (DCA), with differences between the models evaluated using the DeLong test.
Results
Our findings indicate that the clinically combined radiomics model outperforms the standalone radiomics model in prognostic prediction. The area under the curve (AUC) for the combined model was 0.90 in the training cohort, while the validation cohort achieved an AUC of 0.86. DCA further confirmed that the performance of all models exceeded that of blindly predicting patient outcomes as either all negative or all positive. Subsequent comparisons using the DeLong test revealed significant differences in predictive performance between the standalone radiomics model and the clinically combined radiomics model, with P-values <0.05.
Conclusion
The clinically combined radiomics model demonstrates promising performance in predicting the prognosis of NPC patients following IMRT.
扫码关注我们
求助内容:
应助结果提醒方式:
