Decoding network-mediated retinal response to electrical stimulation: implications for fidelity of prosthetic vision

Electrical stimulation of neuroretinas with 3D pyrolytic carbon electrodes

Retinal prosthesis has been one of the medical strategies aimed at restoring some degree of vision for patients affected by retinal degenerative diseases, such as Retinitis Pigmentosa (RP) and age-related macular degeneration (AMD), which are leading causes of irreversible visual loss. In retinal prosthesis, electrical pulses are typically delivered to the retinal neurons via electrodes on the surface of the implant. In this work, we fabricated 3D carbon pillar electrodes by pyrolysis of SU-8 structures defined photolithographically on Si wafers. We then measured compound action potentials induced in porcine neuroretinas stimulated with electrical pulses. The recorded spikes were validated to be biological in origin by adding the voltage-gated sodium-channel blocking agent tetrodotoxin. The minimum threshold voltage needed to effectively stimulate retinal cells, such as retinal ganglion cells, with 3D electrodes was analyzed through systematic investigation of the spike rate and amplitudes as a function of stimulation voltage. 3D electrodes significantly increased spike rate and amplitudes above spontaneous activity in the tissue during stimulation and outperformed the 2D counterpart, both in terms of spike rate and amplitude. Our results indicate a threshold voltage range of 500-600 mV for 1 ms pulses at a frequency of 10 Hz above which a significant increase in spike count was observed. Furthermore, we report an order of magnitude increase in peak-to-peak amplitude for evoked spikes (> 3 mV), compared to spontaneous spikes (∼ 200 µV). Based on numerical integration, we estimate the area under the curve to be ~14 times larger in evoked compound action potentials compared to spontaneous activity. This indicates the relative increase in number of contributing cells to the compound action potential. At a stimulation voltage of 600 mV the spike rate for 3D electrodes was above 10 spikes/channel/s. We hypothesize that the significant difference between 2D and 3D electrodes is not only caused by the higher active electrode surface area of the 3D micropillar electrodes, but also by more intricate contact and interaction with the inner cell layers of the retinal tissue. Our findings indicate that 3D carbon micropillar electrodes are promising for electrical stimulation of the retina.

A Generic Technique to Remove a Variety of Electrical Stimulation Artifacts and Detect Stimulus-Embedded Neural Activity

Investigating neural activity is critical to advancing neuroscience and informing clinically-relevant neuromodulation, but it is often burdened by the presence of recording artifacts resulting from electrical stimulation. We developed and evaluated a generalizable method for removing the electrical artifacts elicited by various neurostimulation waveforms. This investigation of a novel variety of stimulation waveforms was facilitated by obtaining and labelling the neural activity of 15 loose cell-attached patch clamp recordings of retinal ganglion cells. These labelled recordings provided 5785 peristimulus spikes for which the pipeline exhibited a true positive rate of 88.7%, false discovery rate of 2.9%, and d-prime of 3.11, far superseding interpolation (of which bore a d-prime of 0.32). We additionally utilized synthesized spike trains to demonstrate recovery of low-amplitude stimulus-embedded spikes with negligible waveform distortion (low root mean square error). This technique could better enable a rich variety of neurostimulation investigations, and aid in the development of clinical neurostimulation strategies.

Asymmetric Activation of ON and OFF Pathways in the Degenerated Retina

Retinal prosthetics are one of the leading therapeutic strategies to restore lost vision in patients with retinitis pigmentosa and age-related macular degeneration. Much work has described patterns of spiking in retinal ganglion cells (RGCs) in response to electrical stimulation, but less work has examined the underlying retinal circuitry that is activated by electrical stimulation to drive these responses. Surprisingly, little is known about the role of inhibition in generating electrical responses or how inhibition might be altered during degeneration. Using whole-cell voltage-clamp recordings during subretinal electrical stimulation in the rd10 and wild-type (wt) retina, we found electrically evoked synaptic inputs differed between ON and OFF RGC populations, with ON cells receiving mostly excitation and OFF cells receiving mostly inhibition and very little excitation. We found that the inhibition of OFF bipolar cells limits excitation in OFF RGCs, and a majority of both pre- and postsynaptic inhibition in the OFF pathway arises from glycinergic amacrine cells, and the stimulation of the ON pathway contributes to inhibitory inputs to the RGC. We also show that this presynaptic inhibition in the OFF pathway is greater in the rd10 retina, compared with that in the wt retina.

Probing the Contribution of Vertical Processing Layers of the Retina to White-Noise Electrical Stimulation Responses

Optimal stimulus parameters for epiretinal prostheses have been investigated by analyzing retinal ganglion cell (RGC) spiking responses to white-noise electrical stimulation, through a spike-triggered average (STA) analysis technique. However, it is currently unknown as to activation of which retinal cells contribute to features of the STA. We conducted whole-cell patch clamping recordings in ON and OFF RGCs in response to white-noise epiretinal electrical stimulation by using different inhibitors of synaptic transmission in a healthy retina. An mGluR6 agonist, L-AP4, was firstly used to selectively block the output of photoreceptors (PRs) to ON bipolar cells (BCs). We subsequently fully blocked all synaptic inputs to RGCs using a combination of pharmacological agents. Our data shows that PRs dominate the ability of ON RGCs to integrate electrical pulses and form a unique STA shape, while BCs do not contribute in any way. In addition, our results demonstrate that the ability of OFF RGCs to integrate pulses is consistently impaired after blocking the PR to ON BC pathway. We hypothesise that the mechanisms underlying this co-effect are related to the narrow field AII amacrine cells connecting ON and OFF pathways.Clinical Relevance-Recent retinal studies recorded mirror-inverted STAs in ON and OFF retinal pathways, thus raising the possibility of designing a stimulation approach that can differentially activate ON and OFF pathways with electrical stimulation. However, the detailed contribution of three major retinal cell layers in forming characteristic STAs is still unclear. It is of great clinical relevance to investigate the isolated contribution of PRs to the electrically driven STA since PRs progressively degenerate in the course of retinal disease.

Artificial intelligence techniques for retinal prostheses: a comprehensive review and future direction

Objective. Retinal prostheses are promising devices to restore vision for patients with severe age-related macular degeneration or retinitis pigmentosa disease. The visual processing mechanism embodied in retinal prostheses play an important role in the restoration effect. Its performance depends on our understanding of the retina's working mechanism and the evolvement of computer vision models. Recently, remarkable progress has been made in the field of processing algorithm for retinal prostheses where the new discovery of the retina's working principle and state-of-the-arts computer vision models are combined together.Approach. We investigated the related research on artificial intelligence techniques for retinal prostheses. The processing algorithm in these studies could be attributed to three types: computer vision-related methods, biophysical models, and deep learning models.Main results. In this review, we first illustrate the structure and function of the normal and degenerated retina, then demonstrate the vision rehabilitation mechanism of three representative retinal prostheses. It is necessary to summarize the computational frameworks abstracted from the normal retina. In addition, the development and feature of three types of different processing algorithms are summarized. Finally, we analyze the bottleneck in existing algorithms and propose our prospect about the future directions to improve the restoration effect.Significance. This review systematically summarizes existing processing models for predicting the response of the retina to external stimuli. What's more, the suggestions for future direction may inspire researchers in this field to design better algorithms for retinal prostheses.

Differences in the spatial fidelity of evoked and spontaneous signals in the degenerating retina

Vision restoration strategies aim to reestablish vision by replacing the function of lost photoreceptors with optoelectronic hardware or through gene therapy. One complication to these approaches is that retinal circuitry undergoes remodeling after photoreceptor loss. Circuit remodeling following perturbation is ubiquitous in the nervous system and understanding these changes is crucial for treating neurodegeneration. Spontaneous oscillations that arise during retinal degeneration have been well-studied, however, other changes in the spatiotemporal processing of evoked and spontaneous activity have received less attention. Here we use subretinal electrical stimulation to measure the spatial and temporal spread of both spontaneous and evoked activity during retinal degeneration. We found that electrical stimulation synchronizes spontaneous oscillatory activity, over space and through time, thus leading to increased correlations in ganglion cell activity. Intriguingly, we found that spatial selectivity was maintained in rd10 retina for evoked responses, with spatial receptive fields comparable to wt retina. These findings indicate that different biophysical mechanisms are involved in mediating feed forward excitation, and the lateral spread of spontaneous activity in the rd10 retina, lending support toward the possibility of high-resolution vision restoration.

Correlated Activity in the Degenerate Retina Inhibits Focal Response to Electrical Stimulation

Retinal prostheses have shown some clinical success in patients with retinitis pigmentosa and age-related macular degeneration. However, even after the implantation of a retinal prosthesis, the patient’s visual acuity is at best less than 20/420. Reduced visual acuity may be explained by a decrease in the signal-to-noise ratio due to the spontaneous hyperactivity of retinal ganglion cells (RGCs) found in degenerate retinas. Unfortunately, abnormal retinal rewiring, commonly observed in degenerate retinas, has rarely been considered for the development of retinal prostheses. The purpose of this study was to investigate the aberrant retinal network response to electrical stimulation in terms of the spatial distribution of the electrically evoked RGC population. An 8 × 8 multielectrode array was used to measure the spiking activity of the RGC population. RGC spikes were recorded in wild-type [C57BL/6J; P56 (postnatal day 56)], rd1 (P56), rd10 (P14 and P56) mice, and macaque [wild-type and drug-induced retinal degeneration (RD) model] retinas. First, we performed a spike correlation analysis between RGCs to determine RGC connectivity. No correlation was observed between RGCs in the control group, including wild-type mice, rd10 P14 mice, and wild-type macaque retinas. In contrast, for the RD group, including rd1, rd10 P56, and RD macaque retinas, RGCs, up to approximately 400–600 μm apart, were significantly correlated. Moreover, to investigate the RGC population response to electrical stimulation, the number of electrically evoked RGC spikes was measured as a function of the distance between the stimulation and recording electrodes. With an increase in the interelectrode distance, the number of electrically evoked RGC spikes decreased exponentially in the control group. In contrast, electrically evoked RGC spikes were observed throughout the retina in the RD group, regardless of the inter-electrode distance. Taken together, in the degenerate retina, a more strongly coupled retinal network resulted in the widespread distribution of electrically evoked RGC spikes. This finding could explain the low-resolution vision in prosthesis-implanted patients.

Usefulness of Automatic Hyperparameter Optimization in Developing Radiation Emulator in a Numerical Weather Prediction Model

To improve the forecasting accuracy of a radiation emulator in a weather prediction model over the Korean peninsula, the learning rate used in neural network training was automatically optimized using the Sherpa. The Sherpa experiment results were compared with two control simulation results using learning rates of 0.0001 and 1 for different batch sizes (full to 500). In the offline evaluation, the Sherpa results showed significant improvements in predicting longwave/shortwave heating rates and fluxes compared to the lowest learning rate results, whereas the improvements compared to the highest learning rate were relatively small because the optimized values by the Sherpa were 0.4756–0.6656. The online evaluation results over one month, which were linked with the weather prediction model, demonstrated the usefulness of Sherpa on a universal performance for the radiation emulator. In particular, at the full batch size, Sherpa contributed to reducing the one-week forecast errors for longwave/shortwave fluxes, skin temperature, and precipitation by 39–125%, 137–159%, and 24–26%, respectively, compared with the two control simulations. Considering the widespread use of parallel learning based on full batch, Sherpa can contribute to producing robust results regardless of batch sizes used in neural network training for developing radiation emulators.

The impact of synchronous versus asynchronous electrical stimulation in artificial vision

Visual prosthesis devices designed to restore sight to the blind have been under development in the laboratory for several decades. Clinical translation continues to be challenging, due in part to gaps in our understanding of critical parameters such as how phosphenes, the electrically-generated pixels of artificial vision, can be combined to form images. In this review we explore the effects that synchronous and asynchronous electrical stimulation across multiple electrodes have in evoking phosphenes. Understanding how electrical patterns influence phosphene generation to control object binding and perception of visual form is fundamental to creation of a clinically successful prosthesis.