With recent advances in numerical methods and experimen- tal validation, cardiac electrophysiology models can become surrogate tools for improved diagnostics and therapy planning. However, day-to- day clinical applications require models that are accurate and detailed enough to capture the main pathological patterns, but at the same time fast, with near real-time computation time. In particular, the models should be computed in a reasonable amount of time to enable personal- ization and on-line therapy guidance. Towards this goal, we present in this manuscript a novel algorithm adapted to graphics processing units (GPU) that enables near real-time cardiac electrophysiology computa- tion with state-of-the-art cellular models. Our method relies on LBM-EP, a Lattice-Boltzmann method, which is naturally scalable to massively parallel architectures. Tested on a synthetic case and on a patient geom- etry, our experiments demonstrate the high scalability of the algorithm, reaching 10x speed up with respect to the CPU implementation of the algorithm.     «

With recent advances in numerical methods and experimen- tal validation, cardiac electrophysiology models can become surrogate tools for improved diagnostics and therapy planning. However, day-to- day clinical applications require models that are accurate and detailed enough to capture the main pathological patterns, but at the same time fast, with near real-time computation time. In particular, the models should be computed in a reasonable amount of time to enable personal- ization and o...     »