Comparing Classical Ordinary Differential Equation and Neural Network Models for Reduced-Order Single-Cell Electrophysiology

• Yan Barbosa Werneck UFJF
• Bernardo Martins UFJF
• Rodrigo Weber UFJF

DOI: https://doi.org/10.5753/sbcas_estendido.2025.6898

Resumo

Electrophysiology modeling is key for non-invasive diagnostics and understanding heart and brain function. Traditional models use ODEs, from detailed ion channel dynamics to reduced-order phenomenological models. We compare a fast reduced-order model with data-driven and physics-informed neural networks as efficient alternatives to numerical solutions. Using the FitzHugh-Nagumo model, we trained networks with numerical data and model physics, employing architecture optimization, adaptive point density, and time-domain splitting. Inference via TensorRT achieved up to 1.8× speedup over optimized CUDA solvers with minimal accuracy loss. These results showcase neural networks as viable emulators when complexity is controlled.

Referências

Campos, J., Sundnes, J., Dos Santos, R., and Rocha, B. (2020). Uncertainty quantification and sensitivity analysis of left ventricular function during the full cardiac cycle. Philosophical Transactions of the Royal Society A, 378(2173):20190381.

Coorey, Glen, F. G. A. F. D. F. S. V. J. (2022). The health digital twin to tackle cardiovascular disease—a review of an emerging interdisciplinary field. npj Digital Medicine, 5(1):6.

FitzHugh, R. (1961). Impulses and physiological states in theoretical models of nerve membrane. Biophysical journal, 1(6):445–466.

Qian, Sun, Y., Zheng, Wang, L., and Huang, Y. (2022). Accelerating sparse deep neural network inference using gpu tensor cores. In 2022 IEEE High Performance Extreme Computing Conference (HPEC), pages 1–6. IEEE.

Raissi, M., Perdikaris, P., and Karniadakis, G. E. (2017). Physics informed deep learning (part i): Data-driven solutions of nonlinear partial differential equations. arXiv preprint arXiv:1711.10561.

Sel, K., O. D. Z. F. (2024). Building digital twins for cardiovascular health: From principles to clinical impact. Journal of the American Heart Association, 13:e031981.