Computational model of haemodynamics during atrial fibrillation

Atrial fibrillation (AF) is associated with reduced cardiac output, which is correlated with increased symptomatic burden and declined quality of life. Predicting haemodynamic effects of AF remains challenging because of the complex interplay of multiple contributing mechanisms. Computational modelling offers a valuable tool for simulating haemodynamics. However, existing models are lacking the capabilities to both replicate beat-to-beat haemodynamic variations during AF at the same time as being well suited for fitting to clinical data. In the present study, we present a computational model comprising: (1) an electrical subsystem that generates unco-ordinated atrial and irregular ventricular activation times characteristic of AF and (2) a mechanical subsystem that simulates haemodynamics using a reduced order model. The model was fitted to replicate individual haemodynamic measurements from 17 patients in the SMURF study during both normal sinus rhythm (NSR) and AF. The fitted model matched a large majority (75%) of blood pressure and intracardiac pressure measurements in both NSR and AF with absolute simulation errors well below 10 mmHg. Furthermore, a large majority of left atrial and left ventricular ejection fraction measurements during NSR were matched with absolute simulation errors well below 10%. The model consistently underestimated right ventricular diastolic pressure during NSR at the same time as overestimating right ventricular systolic and mean left atrial pressures during AF. The presented approach of modelling atrial activity in AF as unco-ordinated atrial contractions, rather than no atrial contraction, achieved lower overall absolute simulation errors when fitting to individual patients. This computationally efficient model provides a platform for future investigations of patient-specific haemodynamics during AF. KEY POINTS: Atrial fibrillation (AF) is linked to the heart pumping less blood, a higher symptomatic burden and a lower quality of life. Although computational models can help us understand the blood circulation in patients with AF, no current models can both replicate beat-to-beat changes during AF and be fitted to individual patients. We developed a computational model that simulates beat-to-beat haemodynamic changes resulting from the unco-ordinated atrial and irregular electrical activation times characteristic of AF. The computational model was fitted to 17 patients and matched a large majority of arterial and intracardiac pressure measurements and ejection fraction measurements well below 10 mmHg and 10%, respectively. This computationally efficient model provides a platform for future investigations of patient-specific haemodynamics during AF.