Abstract
356
Objectives: To develop a radiomics nomogram to estimate survival in patients with stage III/IV head and neck squamous cell carcinoma (HNSCC) and assess its incremental value to the traditional metabolic parameters for individualized survival prediction.
Methods: This retrospective analysis was approved by the institutional review board. A total of 82 eligible patients with stage III/IV HNSCC, in whom the maximum standard uptake value of tumor (SUVT) and metastatic lymph node (SUVLN) both ≥2.5 on pretreatment 18F-FDG PET/CT, were divided into a training set (n=64) and a validation set (n=18). Radiomics features were extracted from pretreatment PET images of each patient. A radiomics signature was constructed by using the least absolute shrinkage and selection operator (LASSO)-Cox regression model. A radiomics score was calculated to reflect survival probability by using the radiomics signature for each patient. A radiomics nomogram was developed by incorporating the radiomics score and metabolic parameters, including SUVT, SUVLN and lymph node-to-primary tumor uptake ratio (SUVLN/T), by using a multivariate Cox regression model. Nomogram performance was assessed in the training set and validated in the validation set. The incremental value of the PET-based radiomics nomogram to the traditional metabolic parameters for individualized survival prediction was assessed with respect to calibration, discrimination and clinical usefulness.
Results: The radiomics scores and SUVLN were independent factors for survival prediction in patients with stage III/IV HNSCC (both P<0.001, hazard ratio: 22.88 vs.1.17, 95% CI: 5.64-92.71 vs. 1.07-1.28). The radiomics nomogram, consisting of radiomics scores and SUVLN, achieved better prediction efficacy than traditional metabolic nomogram (C-index: 0.83 vs.0.72) with AUC of 0.83 vs.0.74 in the training set, and 0.95 vs. 0.86 in the validation set. A significant difference was observed between the survival curves of the high-risk and low-risk groups in the radiomics nomogram (log-rank P<0.001 in the training set and log-rank P=0.00306 in the validation set). Compared with the traditional metabolic nomogram (log-rank P=0.00981 in the training set and log-rank P=0.227 in the validation set), the radiomics nomogram showed a better discrimination capability.
Conclusions: The PET-based radiomics signature is an independent biomarker for survival prediction in patients with stage III/IV HNSCC. The PET-based radiomics nomogram performed better than the traditional metabolic nomogram for individualized survival prediction in patients with stage III/IV HNSCC, which might assist clinicians in tailoring precise therapy.