Figure 3.

(a): Shap beeswarm plot of logistic regression prediction model. The Shap values of each feature were simply drawn by scatter. Different colors represent the relationship between the eigenvalues and the predicted influence, and the distribution of the eigenvalues is also shown. Red represents high eigenvalues, while blue represents low eigenvalues. (b): Shap global bar plot of logistic regression prediction model. The average value of Shap values of each feature is taken to make a bar chart, which shows the importance of each feature in the prediction. Note: For the convenience of presentation, RadF is the radiomics characteristic code, and its corresponding name is shown in Table 3.