Introduction: Fibrosis creates an arrhythmogenic substrate. Its spatial distribution may be diffuse, patchy, stringy, or a combination of them. The stringy form is notably difficult to implement in cubic meshes, which prompted the design of meshes that align with fiber orientation.
Methods: We developed a 3D bilayer interconnected cable model automatically constructed from the left atrial geometry and epi- and endocardial fiber orientation fields. Discretization was different in the longitudinal (200 µm) and the transverse (100 µm) direction to reflect conduction anisotropy. The advantages of cable models are numerical stability and performance at high resolution, handling of strong anisotropy and interpretation as network of resistors. Diffuse fibrosis was introduced as random uncoupling of 13 to 34% of cell-to-cell longitudinal and transverse connections. Stringy fibrosis was intended to represent collagenous septa and was implemented as a random set of longitudinal lines of transverse uncoupling (along cables) with Poisson-distributed length (mean: 4 mm), for a total of 13 to 31% of the connections being removed. Patterns combining diffuse and stringy fibrosis with matching uncoupling percentages were created. The control case had no uncoupling. Normal propagation and atrial fibrillation were simulated using remodeled Courtemanche kinetics in 5 patterns for each of the three fibrosis types and in the control case.
Results: As compared to control, fibrosis prolonged total activation time by 15% (stringy), 41% (diffuse) and 43% (combined). Conduction slowing was smaller in the stringy case since propagation was mostly longitudinal. During atrial fibrillation, diffuse fibrosis was associated with fragmented reentrant wavefronts with many local conduction blocks. Stringy fibrosis significantly increased anisotropy and stabilized reentrant circuits, while keeping wavefronts relatively smooth. Combined fibrosis caused zigzag propagation as previously demonstrated in microscale models.
Conclusion: Interconnected cable models enable computationally efficient organ-scale high-resolution simulations of microstructure remodeling expressed in terms of longitudinal and transverse uncoupling.