Predicting Long-Term Mortality in TAVI Patients Using Machine Learning Techniques

Figure 1. Computational methods. Schematic workflow for the construction of classification models including feature selection, cross-validation to evaluate the discriminant performance, and resulting interpretation.

Figure 2. Feature selection methods. The most relevant variables identified for each method. MR, mitral regurgitation; ALT, alanine aminotransferase; IVST, interventricular septal thickness; Meangrad, mean aortic pressure gradient; INR, international normalized ratio; PAPS, pulmonary artery systolic pressure; BSA, body surface area; BMI, body mass index; LV, left ventricular; EF, ejection fraction.

Figure 3. Evaluation of mortality prediction for different machine learning (ML) models. Receiver operating characteristic curve from ten-fold cross-validation for mortality prediction. AUC, area under the curve; MLP, multilayer perceptron; GBM, gradient boosting machine; XGBoost, extreme gradient boosting; RF, random forest; LASSO, least absolute shrinkage and selection operator.

Figure 4. Receiver operating characteristic curves for prediction of 5-year mortality: (a) multilayer perceptron vs. logistic regression vs. EuroSCORE II, (b) logistic regression models using different feature selection methods. AUC, area under the curve; LR, logistic regression; GBM, gradient boosting machine; RF, random forest; LASSO, least absolute shrinkage and selection operator.

Figure 5. Permutation feature importance permutation feature importance (PFI) method. More relevant features are associated with more negative values. MR, mitral regurgitation; ALT, alanine aminotransferase; IVST, interventricular septal thickness; Meangrad, mean aortic pressure gradient; INR, international normalized ratio.

Figure 6. Shapley additive explanations value plot. The horizontal axis shows whether the effect of the feature is associated with a higher or lower prediction, while the color indicates whether the value of the feature is high (red) or low (blue) for a given observation.