Cohort Modeling of the Human Atria for in silico Trials

• Applicant: Privatdozent Dr.-Ing. Axel Loewe
• Subject Area: Biomedical Systems Technology ; Cardiology, Angiology ; Medical Physics, Biomedical Technology
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 529821741

Computational models of the atria integrated in multi-scale electrophysiological simulations have provided mechanistic insights into the maintenance and termination of rotational electrical activity occurring during atrial fibrillation. In this context, atrial digital twins aim at replicating a patient's heart in silico for the purpose of personalizing and guiding ablation therapy for an optimal treatment outcome. On the other side, statistical shape models constitute the basis for cohort modeling approaches. They allow deriving multiple geometrical models of the atria that capture the real-world anatomical variability allowing for an entire virtual patient cohort to be compiled that can be further expanded by including functional variability in the simulations. Catheter ablation and anti-arrhythmic drug therapy are the main treatment strategies currently being employed in clinical practice. However, neither option can provide permanent cure for atrial fibrillation and considerable recurrence rates remain. In this work, we will therefore investigate the efficacy of different treatment options through computational modeling applied to large virtual patient cohorts. We will employ large-scale simulations and data mining to identify systematic relations for an optimal choice of the anti- arrhythmic drug and their dose as well as of the location of ablation lines that most likely provide permanent cure for atrial fibrillation for entire subgroups of the population. In this regard, we will compile a virtual cohort of computer models characterized by anatomical variability, cellular heterogeneities and different extents of fibrosis infiltrating the healthy myocardial tissue in the atria. A bi-atrial statistical shape model will serve as a basis to derive a total of 400 anatomical models of the atria with different left and right atrial volumes, pulmonary vein orientations and appendage prominences. Subjecting each atrial model to an arrhythmia induction protocol, we will quantitatively assess each model’s propensity to atrial fibrillation. Furthermore, more than 1 million in silico ECGs in sinus rhythm and during ongoing tachycardia will provide the basis for a machine learning classifier trained to estimate the number of arrhythmia induction points and provide in this way a non-invasive measure for stratifying the risk of atrial fibrillation in each subgroup. The simulated AF episodes will also be the basis to examine the efficacy of different treatment strategies comprising anti-arrhythmic drug therapy and ablation to terminate the arrhythmia and prevent recurrence. The proposed combination of anatomical and functional variability captured in sizeable virtual patient cohorts will allow for numerous further in silico studies in the future.

• DFG Programme: Research Grants
• Co-Investigator: Professorin Dr. Constanze Schmidt