Figure 1

Radiomics study flowchart. Radiomics flowchart. (A) Nodules were manually segmented on plain CT images. (B) Three categories of radiomics features were extracted from original CT, and wavelet features were extracted after wavelet decomposition. (C) After features selection, the most informative radiomics features and clinical features were combined to construct machine learning model. Model performance was assessed using ROC, calibration curve, DCA and et.al.