Prediction of heart arrhythmia onset in patients with ICD using machine learning algorithms

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Cardiac arrhythmia poses a significant risk to patients with implantable cardioverter-defibrillators (ICD).
• Early prediction of arrhythmia onset is crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To predict the onset of cardiac arrhythmia in patients with ICD using machine learning algorithms.
• To determine the earliest possible detection window for arrhythmia events.

Main Methods:

• Utilized R-R interval signals from 28 patients with ICD.
• Extracted 42 signal descriptors and identified relevant features using the Boruta algorithm.
• Applied and compared Random Forest, AdaBoost, XGBoost, LASSO, and SVM machine learning models, focusing on Random Forest performance.
• Assessed prediction accuracy using cross-validation.

Main Results:

• Achieved a maximum Area Under the Curve (AUC) of 0.75 in predicting arrhythmia onset.
• Identified risk groups based on signal characteristics.
• Demonstrated the presence of predictive signals at least 1000 R-R intervals prior to arrhythmia.

Conclusions:

• Predictive signals for cardiac arrhythmia are detectable at least 1000 R-R intervals before the event.
• The strength of these predictive signals increases as the arrhythmia onset approaches.
• Machine learning models, particularly Random Forest, show promise in early arrhythmia detection for ICD patients.