POD-DEIM reduction of computational EMG models

- Accelerate simulations of complex, realistic numerical EMG signals with POD and DEIM.
- Speedup factors of up to 180, while preserving accuracy.
- Possibility to consider a realistic number of motor units in numerical EMG models.

This article presents an application of model order reduction techniques to a numerical model computing electromyographic (EMG) signals. The considered EMG model is a combination of the extracellular bidomain equation with a parameterized nonlinear membrane voltage source. Key ingredients for the proposed reduction methodology are Galerkin projection via proper orthogonal decomposition and application of the discrete empirical interpolation method to the nonlinear source term. The computational efficiency of the approach is demonstrated by numerical examples comparing the reduced model to the full (non-reduced) model. In detail, a high-fidelity model of ≈100 000 degrees of freedom is successfully reduced by several orders of magnitude whilst preserving accuracy and gaining a computational speedup of up to 180. Thus, using the proposed method a realistic number of muscle fibres and motor units can be considered in numerical EMG simulations, which is not feasible using full models due to their high computational cost.