A Nanoradiomics Approach for Differentiation of Tumors Based on Tumor-Associated Macrophage Burden

Figure 1

Pipeline for determination of radiomic signatures. (a) 3D tumor segmentation of n-CECT images. (b) Radiomic features extraction (900 radiomics features: shape, intensity texture based, 3D wavelets). (c) Machine learning: feature selection and reduction of highly correlated features, followed by model training and 5-fold cross-validation for identifying radiomic signature.