Search for: All records

Early-after depolarizations (EADs) are changes in the action potential plateau that can lead to cardiac arrhythmia. At the cellular level, these oscillations are irregular and change from beat to beat due to the sensitivity of voltage repolarization to subcellular stochastic processes. However, the behavior of EADs in tissue, where cells are strongly coupled by gap junctions, is less understood. In this study, we develop a computational model of EADs caused by a reduction in the rate of calcium-induced inactivation of the L-type calcium channel. We find that, as inactivation decreases EADs occur with durations varying randomly from beat to beat. In cardiac tissue, however, gap junction coupling between cells dampens these fluctuations, and it is unclear what dictates the formation of EADs. In this study we show that EADs in cardiac tissue can be modeled by the deterministic limit of a stochastic single-cell model. Analysis of this deterministic model reveals that EADs emerge in tissue after an abrupt transition to alternans, where large populations of cells suddenly synchronize, causing EADs on every other beat. We analyze this transition and show that it is due to a discontinuous bifurcation that leads to a large change in the action potential duration in response to very small changes in pacing rate. We further demonstrate that this transition is highly arrhythmogenic, as the sudden onset of EADs on alternate beats in cardiac tissue promotes conduction block and reentry. Our results highlight the importance of EAD alternans in arrhythmogenesis and suggests that ectopic beats may not be required. 

The electrical signals triggering the heart's contraction are governed by non-linear processes that can produce complex irregular activity, especially during or preceding the onset of cardiac arrhythmias. Forecasts of cardiac voltage time series in such conditions could allow new opportunities for intervention and control but would require efficient computation of highly accurate predictions. Although machine-learning (ML) approaches hold promise for delivering such results, non-linear time-series forecasting poses significant challenges. In this manuscript, we study the performance of two recurrent neural network (RNN) approaches along with echo state networks (ESNs) from the reservoir computing (RC) paradigm in predicting cardiac voltage data in terms of accuracy, efficiency, and robustness. We show that these ML time-series prediction methods can forecast synthetic and experimental cardiac action potentials for at least 15–20 beats with a high degree of accuracy, with ESNs typically two orders of magnitude faster than RNN approaches for the same network size.