The abstract is read at the publisher; the summary is JClub's.
A hybrid biophysical–neural network surrogate model reproduces cardiac action potentials with 1–2 mV RMSE and matches conduction velocity across multiple intercalated disk geometries with 0.24% mean error.
- Why it matters: Resolving intercalated disk nanostructure has been computationally prohibitive, leaving the conduction impact of junction-level remodeling in arrhythmogenic disease largely untestable.
- What they did: The approach trains memory-based flow map surrogates on FEM simulations of single junctions and short cables, then replaces each junction in 2-D/3-D tissue simulations with over one million junctions.
- The result: Surrogates run 10^4–10^5 times faster than FEM and, in large tissues with reduced gap-junctional coupling, regional cleft widening blocks conduction—where either change alone does not.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.