Related Experiment Video
Updated: Jan 8, 2026
Estimating Bilateral Atrial Function by Cardiovascular Magnetic Resonance Feature Tracking in Patients with Paroxysmal Atrial Fibrillation
Published on: July 20, 2022
Atrial fibrillation identification in patients with non-alcoholic fatty liver disease: a machine learning model based
Yao Zhang1, JiaXin Qi2, Xia Chen1
1Department of Cardiovascular Medicine, The Affiliated Hospital of Northwest University Xian, No.3 Hospital, Xian, 710032, Shaanxi, China.
Background:
The global prevalence of non-alcoholic fatty liver disease (NAFLD) is continuously rising, making it a significant public health concern. Recent studies indicate that NAFLD is independently associated with an increased risk of atrial fibrillation (AF), potentially mediated by chronic inflammation and immune responses. However, there is currently a lack of AF risk identification tools specifically for the NAFLD population. This study aimed to identify AF and model its association with immune-inflammatory markers in NAFLD patients.
Methods:
This study enrolled 723 patients with ultrasound-confirmed NAFLD (AF group: n = 203, non-AF group: n = 520). Clinical data were collected, and 10 immune-inflammatory markers (including Systemic Immune-inflammation Index (SII), Systemic Inflammation Response Index (SIRI), Neutrophil-to-Lymphocyte Ratio (NLR), Platelet-to-Lymphocyte Ratio (PLR), Monocyte-to-HDL Cholesterol Ratio (MHR), etc.) were calculated. Feature selection was performed using univariate logistic regression, LASSO, and the Boruta algorithm. Ten machine learning algorithms were employed to construct models, optimized via 10-fold cross-validation. SHapley Additive exPlanations (SHAP) were used to interpret the model, and the best-performing model was ultimately deployed as an online web calculator.
Results:
Multivariate analysis identified SII, NLR, PLR, MHR, and SIRI as independent predictors of AF (all P < 0.05). In the test set, the Support Vector Machine (SVM) model demonstrated the best predictive performance, with an AUC of 0.848 (95% CI: 0.785-0.911). Accuracy, specificity, precision, and F1-score were 0.847, 0.910, 0.745, and 0.713, respectively. SHAP analysis revealed age, PLR, and creatinine as the most influential predictive variables. Based on the final model, we developed a user-friendly online risk calculator.
Conclusion:
This study constructed and validated a machine learning model integrating multiple immune inflammatory markers for effectively identifying the risk of atrial fibrillation in patients with NAFLD. The SVM model exhibited excellent discriminatory ability and clinical applicability. The developed web tool aids in the identification of high-risk individuals and offers new strategies for preventive interventions against AF.
