OBJECTIVE: This research aimed to develop an enhanced CT-based radiomics model capable of predicting CD40LG expression. METHODS: For 399 HNSCC cases, the Cancer Genome Atlas provided gene sequencing, and the Cancer Imaging Archive provided computerized tomography scans. Next, radiomic features were selected through maximum relevance minimum redundancy, and recursive feature elimination. Researchers used gradient boosting machines to develop a radiomics model for CD40LG prediction, evaluating it with operating characteristic curves, decision curves, and the Hosmer-Lemeshow test. RESULTS: The study found that CD40LG levels were significantly lower in HNSCC tissues than healthy ones. High CD40LG expression was associated with better survival rates, as shown by Kaplan-Meier analysis. The findings of regression analyses confirmed that HNSCC patients with higher CD40LG expression had better outcomes. We developed a CD40LG prediction model based on seven key radiomics features. The ROC analysis showed high accuracy for our model, with an AUC of 0.872 for training data and 0.762 for validation data. The calibration and decision curves indicated good model performance in predicting outcomes and applicability. High CD40LG expression was linked to higher RS values, while HNSCC patients with lower RS values fared poorly. CONCLUSIONS: The study concludes that a contrast-enhanced CT-based radiomics model can accurately predict CD40LG expression and prognosis in HNSCC patients, aiding in clinical decision-making. LEVEL OF EVIDENCE: Level 4.
Fan et al. (2026) studied this question.