Radiogenomics Analysis-based Identification of Prognostic Imaging Biomarkers in Hepatocellular Carcinoma

Chen Ying; Yan Zhuangzhi; Gao Xin

doi:10.16289/j.cnki.1002-0837.2017.06.011

Abstract：Objective To identify imaging biomarkers with biological interpretation in hepatocellular carcinoma (HCC) using radiogenomic methodology that associates computed tomography (CT) imaging features with gene modules.Methods Three hundred and seventy one patients who had gene expression profiles from the Cancer Genome Atlas were retrospectively analyzed.For the 37 patients with contrast-enhanced CT imaging data,639 quantitative imaging features were extracted from delineated tumor.Twelve imaging features were selected based on concordance index (CI) criteria.Gene modules acquired from a clustering method were associated with imaging features using Spearman rank correlation;the imaging features that significantly correlated with gene modules were obtained.These imaging features were evaluated with Cox's proportional hazard models to determine their prognostic capabilities.Results Four of 12 imaging features significantly correlated with the prognostic gene modules,and 3 were associated with survival significance.Among these,the wavelet feature Small Zone High Gray-Level Emphasis of wavelet component,which was significantly correlated with the prognostic gene module representing N-Glycan biosynthesis,was predictive of overall survival (OS) (P =0.006,hazard ratio =0.16).The texture feature Long Run Low Gray-Level Emphasis,which was correlated with the prognostic gene module representing angiogenesis,also predicted OS (P =0.049,hazard ratio =3.21).Conclusion Imaging features depicting the wavelet frequency and texture have interpretable biological meaning and can predict overall survival,thereby could be potential imaging biomarkers for HCC patients.