AUTHOR=Xie Yang , Wang Min , Xia Haibin , Sun Huifang , Yuan Yi , Jia Jun , Chen Liangwen 

TITLE=Development and validation of a CECT-based radiomics model for predicting IL1B expression and prognosis of head and neck squamous cell carcinoma

JOURNAL=Frontiers in Oncology

VOLUME=Volume 13 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2023.1121485

DOI=10.3389/fonc.2023.1121485

ISSN=2234-943X

ABSTRACT=Objectives: This study explored the underlying function of IL1B in head and neck squamous cell carcinoma (HNSCC), developed and validated a radiomics model based on contrast-enhanced computed tomography (CECT) to predict IL1B expression, and evaluated its prognostic value in patients with HNSCC.
Methods: A total of 139 patients with RNA-Seq data from The Cancer Genome Atlas (TCGA) and matched CECT data from The Cancer Image Archive (TCIA) were included in the analysis. The prognostic value of IL1B expression in patients with HNSCC was analyzed using Kaplan-Meier analysis, Cox regression analysis and subgroup analysis. Furthermore, the molecular function of IL1B on HNSCC was explored using function enrichment and immunocytes infiltration analyses. Radiomic features were extracted with PyRadiomics and processed using max-relevance min-redundancy, recursive feature elimination, and gradient boosting machine algorithm to construct a radiomics model for predicting IL1B expression. The area under the receiver operating characteristic curve (AUC), calibration curve, precision recall (PR) curve, and decision curve analysis (DCA) curve were used to examine the performance of the model. 
Results: Increased IL1B expression in patients with HNSCC indicated a poor prognosis (hazard ratio [HR] = 1.56, P = 0.003) and was harmful in patients who underwent radiotherapy (HR = 1.87, P = 0.007) or chemotherapy (HR = 2.514, P < 0.001). Shape_Sphericity, glszm_SmallAreaEmphasis, and firstorder_Kurtosis were included in the radiomics model (AUC: training cohort, 0.861; validation cohort, 0.703). The calibration curves, PR curves and DCA showed good diagnostic effect of the model. The rad-score was close related to IL1B (P = 4.490*10-9), and shared the same corelated trend to EMT-related genes with IL1B. A higher rad-score was associated with worse overall survival (P = 0.041). 
Conclusion: The CECT-based radiomics model provides preoperative IL1B expression prediction and offers non-invasive instructions for the prognosis and individualized treatment of patients with HNSCC.