CT Radiomics Combined with Metabolic-Biomarkers Enables Early Recurrence Prediction in Hepatocellular Carcinoma.

Abstract

Background: The prognosis of early recurrence of hepatocellular carcinoma (HCC) remains poor. This study aimed to develop and validate a radiomics model and determine potential biomarkers involved in biological pathways for early recurrence of HCC.

Methods: A total of 271 HCC patients from the First Affiliated Hospital of Guilin Medical University were enrolled as the training cohort. Recurrence related radiomics features were determined by analyzing contrast-enhanced CT images, which were used for the construction of Rad-score. For external validation, we utilized both imaging and transcriptome data from 34 HCC patients in TCGA database. The identified radiomics-related genes were further validated using two independent datasets (OEP000321 and GSE14520) and immunohistochemical analysis of EEF1E1 in 38 HCC tissue samples from training cohort.

Results: Rad-scores based on six radiomics features showed predictive value for early HCC recurrence in both cohorts (A465, A466, A839, V105, V250, V291). Relevant radiomics features are associated with metabolism, proliferation, and immune pathways. The most relevant recurrence-related radiomics gene module was determined via weighted correlation network analysis (WGCNA), which contained LRP12, GPD1L, GARS, EEF1E1, and DGKG. The model based on these genes could efficiently predict early HCC recurrence and was verified in the OEP000321 and GSE14520 datasets. Moreover, EEF1E1 was significantly associated with the Rad-score, illustrated prognostic value at the transcription level, and validated by immunohistochemical staining at the protein level.

Conclusion: Rad-score and radiomics gene signatures from enhanced CT effectively predicted early recurrence in HCC, while EEF1E1 might serve as an efficient biomarker for early recurrence prediction for hepatocellular carcinoma.