Combining the radiomics signature and HPV status for the risk stratification of patients with OPC

Abstract

ObjectiveThe objective was to perform risk stratification of oropharyngeal cancer (OPC) for treatment de‐escalation based on the radiomics analysis and human papillomavirus (HPV) status.MethodsA total of 265 patients with OPC who underwent baseline contrast‐enhanced computed tomography were analyzed, and the patients were grouped into the training (n = 133) and test (n = 132) cohorts at a ratio of 1:1. Intratumoral and peritumoral radiomics features were extracted, and the radiomics signature (Rscore) was calculated using least absolute shrinkage and selection operator regression (LASSO) and Cox regression analyses.ResultsTwelve features were selected to establish the radiomics signature (Rscore) of intratumoral regions +10‐mm peritumoral regions, which yielded maximum AUCs of 0.835, 0.798, and 0.784 in the training, test, and validation cohorts, respectively. Patients with OPC were divided into the high‐risk group, intermediate‐risk group, and low‐risk group based on the Rscore and HPV status and had different prognoses. Patients in the low‐risk group benefit from radiotherapy alone, and patients in the intermediate‐risk group only benefitted from chemoradiotherapy.ConclusionThe radiomics signature appears to improve the predictive performance of clinical characteristics for oropharyngeal cancer. The combined stratification of the radiomics signature and HPV status might be preferred to select patients for de‐escalated treatment.

Risk assessment of venous thromboembolism in head and neck cancer patients and its establishment of a prediction model

“…Lasso algorithm was suitable for shrinking high-dimensional data, especially those showing high levels of multicollinearity, to filter out the optimal predictive features in risk factors from HNC patients with VET and simultaneously avoids overfitting (through regularization). 16 An ordinal logistic regression analysis was subsequently applied to identify the determinants of HNCassociated VTE, deducing a new nomogram model by incorporating the features selected from Lasso method. Predictive value of nomogram model was evaluated via receiver operating characteristic (ROC) curve analysis, calibration curve, and decision curve analysis (DCA).…”

“…R software (https://www.rstudio.com, version 4.2.1, R Foundation for Statistical Computing, Vienna, Austria) was used for mathematical model construction (packages “glmnet,” “pROC,” “rmda,” “rms,” and “caTools”). Lasso algorithm was suitable for shrinking high‐dimensional data, especially those showing high levels of multicollinearity, to filter out the optimal predictive features in risk factors from HNC patients with VET and simultaneously avoids overfitting (through regularization) 16 …”

De-escalated radiation for human papillomavirus virus-related oropharyngeal cancer: evolving paradigms and future strategies

“…The explosion of radiomic information also has the potential to identify who may or may not be eligible for de-escalation, both at diagnosis and midway through radiation. For instance, investigators from China used a radiomics signature of intra-tumoral and peri-tumoral regions to predict which patients might benefit from the addition of chemotherapy to radiation for HPV-related oropharyngeal cancer ( 114 ). Another study showed that radiomics can outperform traditionally used clinical factors to characterize HPV-related oropharyngeal squamous cell carcinoma ( 115 ).…”