Artificial intelligence-enhanced wearable technology enables ventricular arrhythmia prediction.

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Physical behavior patterns may influence ventricular arrhythmia risk.
• Wearable technology offers novel methods for monitoring physical behavior.

Purpose of the Study:

• To investigate the association between temporal dynamics of physical behavior and ventricular arrhythmia risk.
• To evaluate the predictive performance of statistical and deep learning models for ventricular arrhythmia.

Main Methods:

• Prospective study of 303 patients with implantable cardioverter-defibrillators (ICDs).
• Continuous physical behavior data collected via accelerometers over 365 days.
• Analysis using statistical indices and deep neural network representations; logistic regression and cross-validation for prediction.

Main Results:

• Reduced daily physical activity bouts and low sleep variation correlated with higher arrhythmia risk.
• Deep representations yielded improved predictive performance (AUROC 0.74 ± 0.05) compared to statistical indices (AUROC 0.67 ± 0.14).

Conclusions:

• Physical behavior patterns detected by wearables are independently associated with ventricular arrhythmia.
• Deep representations enhance the prediction of ventricular arrhythmia risk.