Figure 11.

Deep learning-based prediction of phase maps and rotor cores or phase singularities (PS) from sparse electrical excitation wave pattern mimicking multi-electrode catheter or optical fiber recordings. (A) Sparse excitation wave pattern with noise (σ = 0.3, 17% coverage). (B) Phase map $\widehat{\varphi }(x,y)$ predicted by neural network analyzing data in (A). (C) Ground-truth phase map ϕ(x, y) obtained with complete, non-sparse, non-noisy data. (D) Spatially resolved angular accuracy (temporal average in each pixel) shows that accuracy decreases between electrodes. (E) Trajectories of ground truth PS (white) and predicted PS (black) using indirect prediction with model M1 (shown over 100 simulation time steps) (see also Supplementary Video 8).