Tutorial: a computational framework for the design and optimization of peripheral neural interfaces

A machine learning framework to optimize optic nerve electrical stimulation for vision restoration

“…The transformation from an imperfect-selectivity stimulation protocol and the corresponding optic nerve activation should come from a less abstract model. The natural choice here would be to insert a hybrid model 35 to compute the response of the optic nerve fibers to electrical stimulation via detailed finite element modeling and neural computation simulation. The setting of an accurate hybrid model for the situation under study would require gaining some further knowledge on the morphology of the optic nerve for the targeted subject population and the definition of viable neuroprosthetic devices.…”

“…Nerve activation in imperfect-selectivity control The activation elicited via imperfect-selectivity control is computed via a hybrid model. 35 Each unit in a filter of the optic nerve layer is associated with a fiber in the true nerve section. We compute in turn the extracellular potential generated by the active sites that stimulates the fiber and the fiber response in terms of an abstract ''firing rate,'' normalized so that it can be used directly as the activation value for the given unit.…”

3D printed biomimetic cochleae and machine learning co-modelling provides clinical informatics for cochlear implant patients

“…The 3D printed biomimetic cochleae offer a robust physical means to capture the variability in anatomical-guided CI insertion, and replicate the ionic conduction and the electron-ion interaction in cochleae with implanted CIs. The 3PNN framework could provide an important tool for solving the 'volume conduction' problem, the first step in computational neuroengineering for modelling electrical stimulation in a biological structure 19 . Further studies that evaluate the correlation between the intracochlear voltage distribution and the excitation of neural cells will be of particular benefit to expand the use of 3PNN in modelling the excitation spread at the neuronal level.…”

A modular strategy for next-generation upper-limb sensory-motor neuroprostheses

“…133 The second category, so far used only for PNS stimulation, can be labeled biomimetic because the pulses are generated according to the output of the simulation of the dynamics of mechanoreceptors. 136,137 One complimentary key addition to algorithms (performed offline or during the training phase) is the simulation of the neural structures receiving the stimulations, usually in combination with finite-element modeling of the electrode structure to optimize the stimulation location/pattern for feedback (see Romeni et al 138 for the PNS and Kumaravelu et al 139 for the CNS). Different sensations have been reported with different encoding strategies, such as force-and-stiffness feedback when using a linear algorithm modulating impulse amplitude, 49 roughness, 140 and texture irregularity, 141 when combined with a biomimetic algorithm generating the pulses with a model of single mechanoreceptor behavior.…”