Predictability as a Marker of Healthy and Pathological ECG

This study examines the complexity of electrocardiographic (ECG) signals, focusing on their short-term predictability. Using predictability measures, 114 ECG recordings from both healthy individuals and cardiac patients are classified. Time series prediction is based on dynamical systems theory, with predictability assessed through the average magnitude and variability of prediction errors. The findings support the hypothesis that cardiovascular diseases may reduce the complexity of heart dynamics. The higher predictability of pathological ECGs compared to healthy ones, along with clustering results, suggests potential clinical applications, such as diagnosing extrasystoles or optimizing resynchronization therapy settings.