Biophysically-accurate and surrogate models for the in silico optimization of peripheral and spinal neuromodulation

Computational models have the potential to help the basic understanding and the clinical use of neuroprosthetic devices, allowing a dramatic reduction in the economic and ethical cost of animal and human experimentation. Complex biophysically accurate models of the electrical stimulation of the nervous system are routinely developed, but their application to drive clinical decisions remains limited. Surrogate models of neural response can help accelerate the standard modelling routines so that personalization and optimization may become computationally feasible. Once an appropriate framework for the modelling of neuromodulation applications is available, advances in one domain can be translated with relative ease to other applications, for example by moving between peripheral and spinal neuroprosthetics. In this talk, we will cover the state of the art of biophysically accurate computational models of neuromodulation and how they can be accelerated and generalized using machine learning and modularity principles.