Fig. 3
From: Multimodal machine learning for 5-year mortality prediction after percutaneous coronary intervention

SHAP summary plot ranking tabular features in the multimodal model. Top features include Ht. (Height), EF (Ejection Fraction), and Sex_F; other important predictors include Creatinine, EXT_DISEASE_1VD, and Age. Colors denote feature value (blue: low, magenta: high); positive SHAP values increase predicted risk.