Figure 2.

Workflow of microsatellite instability (MSI) prediction building and analysis. The tumors were segmented on arterial phase (A, B), delayed phase (C, D) and venous phase (E, F) CT images to form volumes of interest (VOIs). One thousand and thirty-seven quantitative radiomics features were extracted from each patient. The least absolute shrinkage and selection operator (LASSO) was used to select the features. Multivariate logistic regression was used to build radiomics, clinical, and clinicoradiomics combined models for MSI prediction. Finally, the radiomics signature and clinical factors were incorporated into a nomogram for individual evaluation. Receiver operating characteristic curves were used to evaluate the clinical usefulness of the nomogram.