Atrial fibrillation (AF) is associated with an increased risk of heart failure, stroke, and death. Current therapy options, including pharmacological treatment and catheter ablation, are not satisfactory due to high recurrence rates. To generate mechanistic insight that is not only applicable to a single patient but to en-tire subgroups of the population, virtual patient cohorts capturing relevant variability are used. This enables the assessment of different therapy options on a population basis and thus the derivation of specific treatment options for subgroups of the populations.
This project aims to develop methodology for cohort-based in silico studies and use it to evaluate the AF vulnerability of selected patient groups. Therefore, virtual cohorts of the atria will be generated en-compassing anatomical and functional variability.
MR imaging data from UK Biobank will be used to construct a statistical shape model (SSM), featuring different variants regarding age, sex, and AF history. From the atrial SSM, we will derive a large number of atrial geometries (i.e., virtual cohorts) characterized by different atrial shapes, ionic parameters and fibrotic extents as well as their ECGs and make them publicly available for the community. Additionally, the variants of the SSM itself will be published to provide the possibility to generate virtual cohorts featuring certain characteristics for other researchers. The generated cohorts will be validated by comparing features from simulated and clinical 12-lead ECG signals.
With the models derived from the SSM, the identification of subgroups with an increased arrhythmia risk is aimed. To assess the vulnerability non-invasively based on the simulated signals, machine learning techniques will be applied. Moreover, optimal treatment strategies for both, pharmacological treatment and ablation, will be identified by employing the generated virtual cohorts to improve clinical decision-making when treating AF.
