Exploration and validation of radiomics signature as an independent prognostic biomarker in stage III-IVb nasopharyngeal carcinoma

// Fu-Sheng Ouyang 1, * , Bao-Liang Guo 1, * , Bin Zhang 2, 3 , Yu-Hao Dong 4 , Lu Zhang 4 , Xiao-Kai Mo 4 , Wen-Hui Huang 4 , Shui-Xing Zhang 2, 3 and Qiu-Gen Hu 1 1 Department of Radiology, The First People’s Hospital of Shunde, Foshan, Guangdong, P.R. China 2 Medical Imaging Center, First Affiliated Hospital of Jinan University, Guangzhou, Guangdong, P.R. China 3 Institute of Molecular and Functional Imaging, Jinan University, Guangzhou, Guangdong, P.R. China 4 Department of Radiology, Guangdong General Hospital/Guangdong Academy of Medical Sciences, Guangzhou, Guangdong, P.R. China * These authors have contributed equally to this work Correspondence to: Qiu-Gen Hu, email: qiugenhu@126.com Keywords: nasopharyngeal carcinoma, radiomics signature, prognostic, biomarker Received: April 18, 2017     Accepted: June 18, 2017     Published: August 24, 2017 ABSTRACT There is no consensus on specific prognostic biomarkers potentially improving survival of nasopharyngeal carcinoma (NPC), especially in advanced-stage disease. The prognostic value of MRI-based radiomics signature is unclear. A total of 970 quantitative features were extracted from the tumor of 100 untreated NPC patients (stage III-IVb) (discovery set: n = 70, validation set: n = 30). We then applied least absolute shrinkage and selection operator (lasso) regression to select features that were most associated with progression-free survival (PFS). Candidate prognostic biomarkers included age, gender, overall stage, hemoglobin, platelet counts and radiomics signature. We developed model 1 (without radiomics signature) and model 2 (with radiomics signature) in the discovery set and then tested in the validation set. Multivariable Cox regression analysis was used to yield hazard ratio (HR) of each potential biomarker. We found the radiomics signature stratified patients in the discovery set into a low or high risk group for PFS (HR = 5.14, p 0.05). Accordingly, pretreatment MRI-based radiomics signature is a non-invasive and cost-effective prognostic biomarker in advanced NPC patients, which would improve decision-support in cancer care.