Machine Learning Risk Prediction for Incident Heart Failure in Patients With Atrial Fibrillation

Figures

• 

  Figure 1. ROC Curves of 6 ML Models Using All Variables Predictive performance of ML models using all 66 variables was examined for the validation cohort. AUC = area under the curve; GBM = gradient boosting machine; ML = machine learning; ROC = receiver operating characteristic; SVM = support vector machine.

• 

  Figure 2. Top 10 Important Variables According to the SHAP Value Variables are color-coded based on the type of variables (red, background; yellow, past history; green, biomarker and blue; imaging data). CTR = cardiothoracic ratio; GBM = gradient boosting machine; HDL = high-density lipoprotein; LV = left ventricular; NN = neural network; RF = random forest; SHAP = Shapley Additive exPlanation; SVM = support vector machine.

• 

  Figure 3. ROC Curves of the Risk Models for the Validation Cohort (A) Practical ML models using 7 variables. (B) Framingham HF risk model. The 7 variables include age, pre-existing HF, creatinine clearance, cardiothoracic ratio on x-ray, LV ejection fraction, LV end-systolic diameter, and LV asynergy. AUC = area under the curve; GBM = gradient boosting machine; HF = heart failure; LV = left ventricular; ML = machine learning; ROC = receiver operating characteristic; SVM = support vector machine.

• 

  Figure 4. Kaplan-Meier Curves Stratified by Random Forest Algorithm Using 7 Variables The 7 variables include age, pre-existing HF, creatinine clearance, cardiothoracic ratio on x-ray, LV ejection fraction, LV end-systolic diameter, and LV asynergy. CI = confidence interval; HF = heart failure; HR = hazard ratio; ML = machine learning; LV = left ventricular.

• 

  Figure 5. Performance of Practical ML Model Among Patients Without Pre-Existing HF (A) ROC curves of the practical ML models. (B) Kaplan-Meier curves for the incidence of HF hospitalization stratified by the practical ML model. AUC = area under the curve; CI = confidence interval; GBM = gradient boosting machine; HF = heart failure; HR = hazard ratio; ML = machine learning; ROC = receiver operating characteristic; SVM = support vector machine.

• 

  Central Illustration. Machine Learning Prediction for Heart Failure in Atrial Fibrillation Patients Machine learning prediction model was created using over 100 variables included in the data set of the Fushimi AF Registry. The entire cohort data were divided into derivation cohort and validation cohort. Imaging data and biomarkers were nominated as important variables for the prediction of future heart failure events. Finally, 7 variables were selected for practical machine learning model based on validity, feasibility, and applicability from the clinician’s perspective. Our practical machine learning model had a certain level of predictive ability and was able to stratify the risk of hospitalization for heart failure in patients with atrial fibrillation. HF = heart failure; LV = left ventricular.