Abstract:Background: In patients with embolic stroke of undetermined source (ESUS), occult atrial fibrillation (AF) has been implicated as a key source of cardioembolism. However, only a minority acquire implantable cardiac loop recorders (ILRs) to detect occult paroxysmal AF, partly due to financial cost and procedural inconvenience. Without the initiation of appropriate anticoagulation, these patients are at risk of increased ischemic stroke recurrence. Hence, cost-effective and accurate methods of predicting AF in ESUS patients are highly sought after. Objective: We aimed to incorporate clinical and echocardiography data into machine learning (ML) algorithms for AF prediction on ILRs in ESUS. Methods: This was a single-center cohort study that included 157 consecutive patients diagnosed with ESUS from October 2014 to October 2017 who had ILR evaluation. We developed four ML models, with hyperparameters tuned, to predict AF detection on an ILR. Results: The median age of the cohort was 67 (IQR 59–74) years old and the median monitoring duration was 1051 (IQR 478–1287) days. Of the 157 patients, 32 (20.4%) had occult AF detected on the ILR. Support vector machine predicted for AF with a 95% confidence interval area under the receiver operating characteristic curve (AUC) of 0.736–0.737, multilayer perceptron with an AUC of 0.697–0.708, XGBoost with an AUC of 0.697–0.697, and random forest with an AUC of 0.663–0.674. ML feature importance found that age, HDL-C, and admitting heart rate were important non-echocardiography variables, while peak mitral A-wave velocity and left atrial volume were important echocardiography parameters aiding this prediction. Conclusion: Machine learning modeling incorporating clinical and echocardiographic variables predicted AF in ESUS patients with moderate accuracy.

What problem does this paper attempt to address?

The problem that this paper aims to solve is to predict the occurrence of atrial fibrillation (AF) in patients with embolic stroke of undetermined source (ESUS). Specifically, the research objective is to predict atrial fibrillation detected by implantable loop recorder (ILR) using machine learning (ML) algorithms by combining clinical data and echocardiographic data. ### Background and Problem - **Background**: In ESUS patients, occult atrial fibrillation (AF) is considered a key cause of cardiogenic embolism. However, only a small number of patients will receive an implantable loop recorder (ILR) to detect occult paroxysmal AF, mainly due to economic cost and operational inconvenience. If appropriate anticoagulant treatment cannot be initiated in a timely manner, these patients may face a higher risk of recurrent ischemic stroke. - **Problem**: Therefore, there is a need to find an economical and accurate method to predict AF in ESUS patients to reduce the risk of stroke recurrence. ### Research Objectives - **Primary Objective**: Develop a machine - learning prediction model to predict paroxysmal AF detected by ILR in ESUS patients using clinical parameters, biomarkers, and echocardiographic parameters. - **Hypothesis**: The study hypothesizes that a machine - learning model combining clinical and echocardiographic parameters can predict the occurrence of AF with moderate to high accuracy and provide insights into important variables that contribute to this prediction, which may not be identifiable by traditional statistical methods. ### Methods - **Data Source**: The study included 157 consecutive patients diagnosed with ESUS in the stroke unit of a tertiary hospital from October 2014 to October 2017, all of whom underwent ILR evaluation. - **Machine - learning Models**: The study developed four machine - learning models (support vector machine, random forest, extreme gradient boosting, multi - layer perceptron) and improved the prediction performance through hyper - parameter tuning. ### Results - **Baseline Characteristics**: The median age of the study cohort was 67 years (interquartile range 59 - 74 years), and the median monitoring time was 1,051 days (interquartile range 478 - 1,287 days). Among the 157 patients, 32 (20.4%) had AF detected during ILR monitoring. - **Model Performance**: - The area under the curve (AUC) of the support vector machine (SVM) was 0.736 - 0.737. - The AUC of extreme gradient boosting (XGBoost) was 0.697 - 0.697. - The AUC of random forest (Random Forest) was 0.663 - 0.674. - The AUC of multi - layer perceptron (MLP) was 0.697 - 0.708. - **Important Features**: - Age, high - density lipoprotein cholesterol (HDL - C), and heart rate at admission are important non - echocardiographic variables. - Peak mitral A - wave velocity and left atrial volume are important echocardiographic parameters. ### Conclusions - **Conclusion**: A machine - learning model combining clinical and echocardiographic parameters can predict AF in ESUS patients with moderate accuracy. Through this study, the researchers hope to provide a more economical and accurate AF prediction method for ESUS patients, thereby reducing the risk of stroke recurrence.

Machine Learning Modeling to Predict Atrial Fibrillation Detection in Embolic Stroke of Undetermined Source Patients

Machine Learning Models of Postoperative Atrial Fibrillation Prediction After Cardiac Surgery.

Prediction of atrial fibrillation in patients with embolic stroke with undetermined source using electrocardiogram deep learning algorithm and clinical risk factors

Predictors of atrial fibrillation detection in embolic stroke of undetermined source patients with implantable loop recorder

Abstract TP223: Predicting Atrial Fibrillation in Embolic Stroke of Unknown Source

Machine learning prediction of atrial fibrillation in cardiovascular patients using cardiac magnetic resonance and electronic health information

Artificial intelligence predicts undiagnosed atrial fibrillation in patients with embolic stroke of undetermined source using sinus rhythm electrocardiograms

Predictors of atrial fibrillation after embolic stroke of undetermined source in patients with implantable loop recorders

Machine learning model for predicting left atrial thrombus or spontaneous echo contrast in non-valvular atrial fibrillation patients based on multimodal echocardiographic parameters

Prediction of atrial fibrillation and stroke using machine learning models in UK Biobank

Short-term atrial fibrillation detection using electrocardiograms: A comparison of machine learning approaches

Machine Learning Algorithm to Predict Atrial Fibrillation Using Serial 12-Lead ECGs Based on Left Atrial Remodeling

Causative Classification of Ischemic Stroke by the Machine Learning Algorithm Random Forests

Using Machine Learning to Predict Atrial Fibrillation Diagnosed after Ischemic Stroke.

Atrial fibrillation detection on implantable loop recorders in embolic stroke of undetermined source

Optimization of Using Multiple Machine Learning Approaches in Atrial Fibrillation Detection Based on a Large-Scale Data Set of 12-Lead Electrocardiograms: Cross-Sectional Study

Predicting stroke in Asian patients with atrial fibrillation using machine learning: A report from the KERALA-AF registry, with external validation in the APHRS-AF registry

Identifying at-risk patients for atrial high-rate episodes by machine learning

Machine learning for outcome prediction of atrial fibrillation recurrence after catheter ablation

Machine learning for stroke in heart failure with reduced ejection fraction but without atrial fibrillation: A post‐hoc analysis of the WARCEF trial

Development and Validation of Machine Learning Algorithms to Predict 1-Year Ischemic Stroke and Bleeding Events in Patients with Atrial Fibrillation and Cancer