Figure 6. The flowchart of radiomics analysis and omics integration. (1) Manual delineation of tumor region of interest (ROI) of contrast-enhanced CT. Shape-based, first-order and second-order radiomics features of ROIs were then calculated. (2) Prediction of somatic mutations and molecular subtypes using radiomics features and multiple machine learning algorithms in independent training/test sets. (3) Radiomics, genomics, transcriptomics and proteomics were integrated to build predictive models for overall survival in training set, and their prognostic values were estimated using validation set.