Synthetic cardiac electrophysiology data from the paper: Latent PDE mapping for efficient physics-informed learning across geometries with limited data

Dataset Description This dataset contains finite element data generated by solving the Aliev-Panfilov PDE using openCARP on geometrically deformed computational meshes. Data Format Each dataset family is stored as a single HDF5 (.h5) file. The file contains one group per geometry, indexed 0, 1, ..., N-1. Each group stores the following arrays: Field Description x Deformed mesh node coordinates, shape (n_nodes, d) x_elems Element connectivity of the deformed mesh x_ref Reference (latent) mesh node coordinates, shape (n_nodes, d) x_ref_elems Element connectivity of the reference (latent) mesh Vm Transmembrane voltage time series, shape (n_nodes, n_timesteps) W Recovery variable time series, shape (n_nodes, n_timesteps) t Simulation times in ms, shape (n_timesteps,) F Deformation gradient tensor at each node, shape (n_nodes, d*d) D Local anisotropic conductivity tensor at nodes, shape (n_nodes, d, d) stim_indices Indices of stimulus nodes pca_nodes Uniformly sampled nodes from a structured mesh mapped through the same transformation, shape (n_pca_nodes, d) F_compute_time Computational time used to calculate the deformation gradient F between the latent geometry and the deformed geometry The groups also include arrays that are deliberately set to None or as an empty array. These arrays are named x_bc, x_ref_bc, x_bc_normals, x_ref_bc_normals, D_bc, F_bc, Vm_bc, org_verts, and downsample_factor.