Fig. 2

Univariable logistic regression analyses to identify predictors of Grade B/C PHLF in patients with HBV-related HCC in the training cohort. Forest maps show the risk ratios of indicators. b Correlation analysis among indicators significantly related with grade B/C PHLF by logistic univariate analysis. Colors from red to blue indicate a correlation from positive to negative. The values represent the significant P values of the correlations, indicating the parts of correlations are significant. c The importance of the Stochastic Forest algorithm based on grouping indexes. Logistic univariate significant indicators were divided into seven groups according to clinical significance and a random forest model was constructed for each group of indicators to predict grade B/C PHLF risk. The bars represent the importance of each indicator; the red bars represent the most important indicators of each group. d There is no correlation among the indicators after redundancy removal by grouping stochastic forest algorithm. Colors from red to blue indicate a correlation from positive to negative. The values inside the circle represent the significant P values of the correlations, indicating the correlations among all indicators are not significant