Bayesian inference for biophysical neuron models enables stimulus optimization for retinal neuroprosthetics

While multicompartment models have long been used to study the biophysics of neurons, it is still challenging to infer the parameters of such models from data including uncertainty estimates. Here, we performed Bayesian inference for the parameters of detailed neuron models of a photoreceptor and an OFF- and an ON-cone bipolar cell from the mouse retina based on two-photon imaging data. We obtained multivariate posterior distributions specifying plausible parameter ranges consistent with the data and allowing to identify parameters poorly constrained by the data. To demonstrate the potential of such mechanistic data-driven neuron models, we created a simulation environment for external electrical stimulation of the retina and optimized stimulus waveforms to target OFF- and ON-cone bipolar cells, a current major problem of retinal neuroprosthetics.

Fractal Electrodes as a Generic Interface for Stimulating Neurons

The prospect of replacing damaged body parts with artificial implants is being transformed from science fiction to science fact through the increasing application of electronics to interface with human neurons in the limbs, the brain, and the retina. We propose bio-inspired electronics which adopt the fractal geometry of the neurons they interface with. Our focus is on retinal implants, although performance improvements will be generic to many neuronal types. The key component is a multifunctional electrode; light passes through this electrode into a photodiode which charges the electrode. Its electric field then stimulates the neurons. A fractal electrode might increase both light transmission and neuron proximity compared to conventional Euclidean electrodes. These advantages are negated if the fractal’s field is less effective at stimulating neurons. We present simulations demonstrating how an interplay of fractal properties generates enhanced stimulation; the electrode voltage necessary to stimulate all neighboring neurons is over 50% less for fractal than Euclidean electrodes. This smaller voltage can be achieved by a single diode compared to three diodes required for the Euclidean electrode’s higher voltage. This will allow patients, for the first time, to see with the visual acuity necessary for navigating rooms and streets.

Neural stimulation for visual rehabilitation: advances and challenges

Blindness affects tens of million people worldwide and its prevalence constantly increases along with population aging. In some pathologies leading to vision loss, prosthetic approaches are currently the only hope for the patient to recover some visual perception. Here, we review the latest advances in visual prosthetic strategies with their respective strength and weakness. The principle is to electrically stimulate neurons along the visual pathway. Ocular approaches target the remaining retinal cells whereas brain stimulation aims at stimulating higher visual structures directly. Even though ocular approaches are less invasive and easier to implement, brain stimulation can be applied to diseases where the connection between the retina and the brain is lost such as in glaucoma and could therefore benefit to patients with different pathologies. Today, numbers of groups are investigating these strategies and the first devices start being commercialized. However, critical bottlenecks still impair our scientific efforts towards efficient visual implants. These challenges include electrode miniaturization, material optimization, multiplexing of stimulation channels and encoding of visual information into electrical stimuli.

Visual cortex responses to suprachoroidal electrical stimulation of the retina: effects of electrode return configuration

A clinically effective retinal prosthesis must evoke localized phosphenes in a retinotopic manner in response to stimulation of each of the retinal electrodes, evoke brightness cues over a wide dynamic range and function within safe stimulus limits. The effects of varying return configuration for retinal stimulation are currently unknown. To investigate this, we implanted a flexible, 7 × 12 electrode array into the suprachoroidal space of normally-sighted, anesthetized cats. Multi-unit activity in the primary visual cortex was recorded in response to electrical stimulation using various return configurations: monopolar vitreous (MPV), common ground (CG), hexagonal (HX), monopolar remote (MPR) and bipolar (BP_N). MPV stimulation was found to be the most charge efficient and was most likely to induce cortical activity within safe charge limits. HX and CG stimulation were found to exhibit greater retinal selectivity compared to the MPV return at the expense of lower cortical yield and higher P50 charge levels, while cortical selectivity was unaffected by choice of return. Responses using MPR and widely spaced BP_N configurations were similar to those using the MPV return. These results suggest that choice of return configuration for a retinal prosthesis will be balanced between resolution and stimulation within safe charge limits.

Calcium channel dynamics limit synaptic release in response to prosthetic stimulation with sinusoidal waveforms

Extracellular electric stimulation with sinusoidal waveforms has been shown to allow preferential activation of individual types of retinal neurons by varying stimulus frequency. It is important to understand the mechanisms underlying this frequency dependence as a step toward improving methods of preferential activation. In order to elucidate these mechanisms, we implemented a morphologically realistic model of a retinal bipolar cell and measured the response to extracellular stimulation with sinusoidal waveforms. We compared the frequency response of a passive membrane model to the kinetics of voltage-gated calcium channels that mediate synaptic release. The passive electrical properties of the membrane exhibited lowpass filtering with a relatively high cutoff frequency (nominal value = 717 Hz). This cutoff frequency was dependent on intra-axonal resistance, with shorter and wider axons yielding higher cutoff frequencies. However, we found that the cutoff frequency of bipolar cell synaptic release was primarily limited by the relatively slow opening kinetics of L- and T-type calcium channels. The cutoff frequency of calcium currents depended nonlinearly on stimulus amplitude, but remained lower than the cutoff frequency of the passive membrane model for a large range of membrane potential fluctuations. These results suggest that while it may be possible to modulate the membrane potential of bipolar cells over a wide range of stimulus frequencies, synaptic release will only be initiated at the lower end of this range.

Electric crosstalk impairs spatial resolution of multi-electrode arrays in retinal implants

Active multi-electrode arrays are used in vision prostheses, including optic nerve cuffs and cortical and retinal implants for stimulation of neural tissue. For retinal implants, arrays with up to 1500 electrodes are used in clinical trials. The ability to convey information with high spatial resolution is critical for these applications. To assess the extent to which spatial resolution is impaired by electric crosstalk, finite-element simulation of electric field distribution in a simplified passive tissue model of the retina is performed. The effects of electrode size, electrode spacing, distance to target cells, and electrode return configuration (monopolar, tripolar, hexagonal) on spatial resolution is investigated in the form of a mathematical model of electric field distribution. Results show that spatial resolution is impaired with increased distance from the electrode array to the target cells. This effect can be partly compensated by non-monopolar electrode configurations and larger electrode diameters, albeit at the expense of lower pixel densities due to larger covering areas by each stimulation electrode. In applications where multi-electrode arrays can be brought into close proximity to target cells, as presumably with epiretinal implants, smaller electrodes in monopolar configuration can provide the highest spatial resolution. However, if the implantation site is further from the target cells, as is the case in suprachoroidal approaches, hexagonally guarded electrode return configurations can convey higher spatial resolution.

Simulations to study spatial extent of stimulation and effect of electrode-tissue gap in subretinal implants

We present a finite element based simulation and analysis method to describe the spatial extent of stimulation and the effects of electrode-tissue interactions in subretinal prostheses. In particular, we estimate the threshold stimulation current needed to depolarise and evoke action potentials in the ganglion cells to be stimulated at a particular distance from the electrode. This is achieved through the application of a threshold electric field to a spherical neuronal soma model of a retinal ganglion cell under consideration. Threshold stimulation currents and the lateral extent of the stimulation zone were computed for disc microelectrodes in subretinal stimulation mode. Recent evidence indicates a decrease in threshold charge with time following subretinal implantation. Consequently, to explain the variation in threshold stimulation currents, we propose a hypothesis based on an electrode-tissue gap. Threshold stimulation currents and impedances for different electrode-tissue gaps were computed. We validate the hypothesis with our simulation results that the changes in impedance observed with time in vivo can be mainly attributed to the varying distance of the ganglion cells from electrodes due to changes in electrode-tissue gap. Our simulation framework proposes a convenient and practical method applicable for studying different electrode geometries used for retinal stimulation.