MRI delta radiomics plus clinical data with XGBoost predicts nasopharyngeal carcinoma chemotherapy response

XGBoost model using MRI delta radiomics and clinical parameters predicted short-term chemotherapy response in nasopharyngeal carcinoma with external validation AUC 0.926 (training AUC 0.962, CI 0.937-0.998).

• Independent predictors of response included T stage (OR 4.29), lymphocyte count (OR 2.08), LDH (OR 1.02), Dynamic Rad_Score (OR 0.65), and MTPC score (OR 0.33).
• SHAP analysis identified Dynamic Rad_Score, MTPC, LDH, and T stage as the top four features.
• Study was multicenter with internal training/test (n=151/65) and external validation (n=87) cohorts.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.