Optimisation of the spatial discretisation of myelinated axon models

Sensory feedback in assistive prosthetic devices is a promising method to improve the quality of life of patients after amputation. However, the complexity of providing a truly natural thermal and mechanical feedback to the nervous system remains a challenge for the future generation of prostheses. Investigations for such a technological progress strongly rely on the modelling of the interfaces of the prostheses with the peripheral nervous system. Such models have to accurately mimic the physiological response of the tissue. Because these models are solved by computational methods, one of the keypoints for high accuracy is the spatial discretisation or meshing. In this paper we propose addressing the discretisation of myelinated axons, taking in consideration the existing rules for other neural structures. Our approach takes into account both the accuracy and the computational cost. We conducted simulations over a wide range of discretisation choices to quantify the deviation of the neural signal propagation velocity. Results showed that one segment per node of Ranvier is enough to model the active sites of a myelinated axon with sufficient accuracy. Modeling internodal myelinated regions (IN), on the other hand, requires 5–15 segments per region for accurate results in most cases, depending on the size of the IN. Our results show that simple guidelines can be followed to significantly reduce the errors in simulations of nerve bundles. Such performances will enable the simulation of more complex phenomena in a context of models associated with chaotic dynamics.