Computerized tomography-based first order tumor texture features in non-small cell lung carcinoma treated with concurrent chemoradiation: A simplistic and potential surrogate imaging marker for survival.

Abstract

This study examines the role of tumor texture on computed tomography (CT) images as a complement to clinical prognostic factors in predicting survival in patients of non-small cell lung carcinoma (NSCLC) treated with radical chemo-radiation (CRT). A total of 93 patients with confirmed NSCLC treated with CRT accrued in a study approved by the institutional ethics committee were analyzed for CT-based radiomic features. Pretreatment CT images were used to contour the primary tumor and texture features were computed by the image filtration method to differentially highlight fine to coarse textures. Texture parameters included mean intensity, entropy, kurtosis, standard deviation, and mean positive pixel and skewness. Optimal threshold cut-off values of the above tumor texture features were analyzed. These features were explored as imaging biomarkers to predict survival using Kaplan-Meier and Cox proportional hazard model. Median follow-up of the entire cohort was 23.5 months [Interquartile range, IQR: 14-37] while for alive patients, median follow-up was 31 months (IQR: 23-49), 47 (50.6%) patients had died at the last follow-up. Univariate analysis revealed certain features like age, gender, response to therapy, and texture features like mean and kurtosis in CT images to be significant predictors of survival. In multivariate analysis, age (P = 0.006), gender (P = 0.004), treatment response (P< 0.0001), and two CT texture parameters: mean (P = 0.027) and kurtosis (P= 0.002) were independent prognostic factors of survival. CT-derived tumor heterogeneity (mean and kurtosis) complements clinical factors for predicting survival in NSCLC patients treated with CRT. Tumor radiomics warrants further validation as potential prognostic biomarkers for these patients.