Computational Modeling and Analysis of Electrophysiology and Hemodynamics During Atrial Fibrillation

Atrial fibrillation (AF) is the most common arrhythmia globally and is characterized by uncoordinated atrial activation and ineffective atrial contraction. Atrial fibrillation is associated with an impaired quality of life and an increased risk of stroke, heart failure, and death. Changes in the autonomic nervous system (ANS) control of cardiac function are a known pathophysiological mechanism in AF, but methods for estimating autonomic activity during AF are lacking. Moreover, although AF is primarily an electrophysiological disease, its detrimental effects are mainly due to its hemodynamic consequences and are challenging to predict. Computational modeling offers a mechanistic framework that holds the potential to address both of these issues. Hence, this thesis aimed to (1) develop and use a computational model of the AV node to study patient-specific ANS modulation on AV nodal conduction properties during AF based on electrocardiogram (ECG) recordings and (2) develop and use a computational model to study patient-specific hemodynamics during AF based on ECG recordings and hemodynamic measurements. In addition, although f-wave characteristics derived from ECG are indicators of atrial remodeling, their prognostic value has not been explored in early-stage AF. The third aim of this thesis was to (3) study the prognostic value of f-wave characteristics in ECG recordings from implantable loop recorders at the earliest stages of AF.To address the first aim, an AV node model from a previously published formulation was extended to incorporate ANS-induced changes in AV nodal refractoriness and conduction delay. Paper I demonstrated the necessity of accounting for ANS modulation of the AV node to accurately replicate observed changes in heart rate variability during tilt tests. The AV node model was further refined by incorporating respiration-induced autonomic modulation. The resulting model generated training data for a convolutional neural network (CNN) to estimate respiration-induced autonomic modulation of the AV nodal function from ECG data. Paper II showed that the developed CNN could effectively estimate respiration-induced autonomic modulation from ECG data, suggesting its potential for monitoring changes and detecting individual differences. However, further validation with ground-truth ANS data is required. The second aim was addressed by developing a computational model that combines an electrical subsystem, including the refined AV node model, with a mechanical subsystem describing cardiovascular mechanics to predict AF-induced hemodynamic changes. Paper III successfully replicated patient-specific arterial and intracardiac pressures using the integrated hemodynamic model, although discrepancies in right ventricular diastolic pressure indicated a need for further model refinement. The third aim was addressed by analyzing f-wave characteristics from ECG recordings of a large cohort of patients with early-stage AF to assess their prognostic value. Paper IV revealed that lower f-wave indices (atrial fibrillatory rate, organization index derived from the signal spectral characteristics, and average amplitude of the f-wave envelope) in early-stage AF have potential as prognostic markers for increased totaland cardiovascular mortality in patients with AF episodes lasting ≥60 minutes. In conclusion, the thesis presents several contributions to developing computational models and analytical methods for understanding and managing AF. The development of patient-specific models of AV nodal conduction and hemodynamics, combined with the investigation of prognostic f-wave characteristics, offers potential pathways toward more personalized and effective treatment strategies for AF.