Figure 1.

A flowchart showing the radiomics analysis for MVI prediction. The clinicoradiological characteristics (especially the status of MVI) were identified first. ROI segmentation was performed on axial LAVA MR images, and then radiomic features were extracted, including shape features, first-order features, textural features, and wavelet transformed features. Next, features with high stability (ICC > 0.8) were included and further selected via mRMR, LASSO and stepwise regression analysis with AIC. The MVI prediction model was constructed by incorporating the radiomics signature and clinicoradiological risk factors. A nomogram was adopted to present the model and evaluated with calibration curve and decision curve analysis.