oa Electrode and Stimulation Parameters for a Thalamic Visual Prosthesis

Introduction

Our long term goal is to develop a visual prosthesis which interfaces with the thalamic visual center of the brain to restore vision to patients who have lost sight through glaucoma, the leading cause of blindness for which no satisfactory medical treatment exists. Glaucoma is a group of neurodegenerative disorders that affects the retinal ganglion cell (RGC) layer of the eye in 60 million people worldwide [1] and is a disease especially prevalent among peoples of the Gulf region. RGCs link the early visual signal processing of the eye with subsequent visual processing centers of the brain. Of these structures, the lateral geniculate nucleus (LGN) located in the thalamus is the primary target for RGC signals. Many groups are seeking to combat vision loss through use of prostheses; currently, two retinal prostheses are approved clinical devices. Because these retinal prostheses require intact RGCs, they have merit for patients with diseases like age–related macular degeneration (AMD) and retinitis pigmentosa, but not for patients who face blindness through glaucoma. Our proposal that seeks to address all people rendered blind, either through diseases of photoreceptor degeneration (e.g., AMD) or through loss of RGCs (i.e., glaucoma) is to develop a high density multi–electrode array (MEA) that can be used to electrically stimulate many neurons of the human LGN (Fig. 1). Figure 1. Visual processing begins in the eye then data are relayed into the brain via the optic nerve to the LGN and subsequently to the visual cortex. Although each of these structures has been a target for visual prostheses, we propose that MEA stimulation of the LGN is a promising technique to address all forms of blindness.

Objectives

The primary objective of the current study has been to determine the optimum electrode materials and stimulus parameters to safely and efficiently stimulate LGN neurons electrically using a multielectrode array (MEA).

Methods

Retina preparation

The retina is a displaced part of the central nervous system that develops like the thalamus from the embryonic diencephalon. RGCs therefore provide for a convenient and advantageous target, because they can be both stimulated visually and electrically in vitro, for early stage testing of electrode performance for LGN stimulation. Retinal tissue was taken from 3–6 month old C57BL/6 wild–type mice. To maintain normal visual–evoked responses from RGCs and be assured that the retina is undamaged before testing the electrodes, mice were sacrificed without drugs and retinas were isolated in darkness. Briefly, mice were first dark–adapted for 1 hr to retain light sensitivity then euthanized by cervical dislocation. Eyes were immediately removed and the retinas were excised in artificial cerebrospinal fluid (ACSF) within 15–20 min using an IR stereomicroscope. The retina was flattened on the MEA with the ganglion cell layer adjacent to the electrodes. The retina was maintained at room temperature (21°C) and perfused with oxygenated ACSF (95% O2, 5% C02) at 3–5 ml/min. Visual and electrical stimulation The MEAs (Multichannel Systems, Reutlingen, Germany) used in experiments consisted of 60 electrodes (30 μm in diameter, separated by 200 μm) arranged in a 2–D planar 8 × 8 rectangular array with a glass ring fixed to the surface to create a chamber for the perfusion medium. The three electrode coatings tested were titanium nitride (TiN), iridium oxide (IrOx), and poly (3,4–ethylenedioxythiophene)–carbon nanotubes (PEDOT–CNT). Visual stimulation of the retina with Gaussian white noise light patterns for 30 min at a 15 Hz refresh rate was performed utilizing a lens to focus the image of a LCD microdisplay (Kopin Ruby SVGA Microdisplay, Edmund Optics, Barrington, NJ, USA) onto the same plane as the MEA electrodes [2]. Electrical stimulation and recording of neurons was performed in a MEA system (STG1002 and MEA1060–Inv–BC, Multichannel Systems, Reutlingen, Germany). The STG1002 stimulator was programmed to deliver current–controlled biphasic symmetrical waveforms with positive or negative phase first, square or sine waveforms, current amplitude ± 1–500 μA, single phase pulse width 40–1000 μs, and frequency 1–200 Hz. Spike Sorting RGC spike trains were processed using a combination of MC Rack to record raw spike records (Multichannel Systems, Reutlingen, Germany), Offline Sorter to identify individual unit spike waveforms (v3 spike sorter, Plexon Inc., Dallas, TX, USA), Matlab for further sorting of units (Mathworks Inc., Natick, MA, USA), and Neuroexplorer for data analysis (v3, Nex Technologies, Madison, AL, USA). The records were sampled at 10 kHz after high pass filtering at 200 Hz. The inherent noise of the system was 5–20 μV depending upon the electrode material, and the voltage threshold for a spike event was set at 20–24 μV to ensure all spikes were at or above the noise level.

Results

Electrode material sensitivity for recording neural spikes Filtered recordings from the three electrode materials were compared to determine their sensitivities for stimulation of RGCs. Several steps were used to ensure recordings obtained were real spikes and not noise or artifacts from the delivered electrical stimulation. Firstly, stimulation signals are biphasic, meaning the charge delivered is equal to the charge returned, eliminating residual charge in the area. The MEA system employs a blanking circuit by which the recording electrodes become disconnected from the MEA during stimulation and do not reconnect until 640 μs has passed in order to avoid electrical artifacts. Through MC Rack software, a 200 Hz high pass filter was implemented so all slow–rising spikes are removed and spike thresholds were set above the noise level for each material. Finally, spikes were sorted with Offline Sorter to separate spikes that are neuronal units from artificial “neuronal like” spikes. The graphs in Fig. 2 illustrate the filtered analog recordings (raw data without threshold, before sorting) of the three electrode materials when stimulating retinas with a 1 μA, 200 μs, 50 Hz biphasic square pulse. The data indicate that recordings with TiN electrodes recorded more noise (noise level around 10–12 μV) than IrOx or PEDOT–CNT electrodes (noise level 5–8 μV). Although the higher noise might suggest that TiN electrodes may be a more sensitive material for recording neural spikes, in fact the signal–to–noise ratio (SNR) was worse than for electrodes of the other two materials, making it less desirable for use as a recording electrode. TiN appeared to record more spikes in the analog data (Fig. 2), however further analysis (as shown later) identified the spikes as being stimulus artifacts. Because TiN delivers stimuli through capacitive charging, it is possible that the blanking time is insufficient and residual charge results in false spikes. IrOx and PEDOT–CNT provided better recording performance with stronger SNR and more reliable spikes. Figure 2. The plots above show representative examples of the raw records obtained from the MEA system for stimulation through different electrode materials. The records come from an electrode 600 μm from the stimulating electrode. The y–axis is the recorded signal amplitude (μV) and the x–axis is a section of time during a stimulus train (0–350 ms). The timing of electrical stimulation pulses (50 Hz) is shown as black bars above the plots. Stimulation efficacy for different electrode materials Tests were conducted on two mouse retinas for each electrode material. Electrode 55 (column 5, row 5) was assigned to be the stimulation electrode and a ground electrode was placed in the media above the MEA. Electrical stimuli were delivered with either a rectangular or a sine wave shape while changing one stimulus parameter at a time; the standard stimulus was a constant 50 μA, 200 μs, 100 Hz. Each stimulus waveform was delivered to the retina 2000 times. Recordings were analyzed for each electrode from − 2 to ?8 ms surrounding a stimulus. The heatmap (Fig. 3) displays the total number of cell spikes recorded during this time window arranged in an 8 × 8 array corresponding to the MEA. Results for two electrode materials are shown (PEDOT–CNT and TiN). The results for IrOx were similar to those for TiN. For the same stimulus conditions, we found that stimulation with PEDOT–CNT electrodes was considerably more effective in evoking spike responses from mouse RGCs than was the case with either stimulation with IrOx or TiN electrodes. Increasing the amplitude of the current used for stimulation through PEDOT–CNT electrodes increased spike responses but not for stimulation with TiN or IrOx electrodes in the range that we tested. Waveform shape, frequency, and pulse width did not appear to affect the number of spikes after stimulation much. The fact that evoked responses were insensitive to increases in stimulus amplitude through IrOx or TiN electrodes suggests that we may have failed to reach the threshold to drive many RGCs for stimulation with these materials. In any event, the key result is quite clear – PEDOT–CNT electrodes provide a far superior interface for electrical stimulation of RGCs and, by inference, LGN neurons. Figure 3. Heat maps showing the total number of spikes recorded within − 2 to ?8 ms of an electrical stimulus at each position in an 8 × 8 array correlating to the MEA electrode layout. Dark blue ? 0 spikes and Red–brown ? 3000 spikes recorded during a period of 2000 stimuli. By increasing the current amplitude from 50–500 μA, more spikes were recorded when stimulation was fed through PEDOT–CNT electrodes, but no significant increase was observed when stimulation was through TiN or IrOx (not shown) electrodes. Square or sine wave shape did not appear to affect the degree of RGC stimulation much. Comparing visual and electrical stimulation Both visual and electrical stimulation can evoke responses from RGCs (Fig. 4) but there are some differences. The main difference we observed was that multi–unit responses were more frequent during visual stimulation. This is likely because axonal units with receptive fields and thus axon initial segments (the point of low threshold spike activation) distant from the electrical stimulation site could be activated by visual but not electrical stimulation and recorded on the electrode of interest. This hypothesis does however require further investigation. Current results suggest that somal units recorded on an electrode can be equally activated by either electrical or visual stimulation, as shown in the plots below where the average waveform shape are nearly identical. Figure 4. The plots above show the set of spike shapes recorded 600 μm from the stimulating electrode. The green lines represent all the waves recorded from a single channel. Red is the average waveform (template) and blue is ± 3 standard deviations. (A) Waveforms were recorded during (A) visual stimulation and (B) electrical stimulation. There are fewer spikes recorded during electrical stimulation but the population averages are not significantly different.

Conclusion

We have tested three different materials for recording and electrical stimulation of RGCs: TiN, IrOx, and PEDOT–CNT. The IrOx and PEDOT–CNT electrodes both have lower noise compared with the TiN electrode. Most significantly, the PEDOT–CNT electrode performed considerably better in stimulating RGCs electrically than both of the other two electrode materials. Changing the current amplitude elicited changes in spike recordings more so than any changes in duration, frequency, or waveform. We also found electrical stimulation can evoke spikes as readily as visual stimulation, indicating that electrical stimulation may indeed interface naturally with the visual system. The major result of this study is that electrodes coated with PEDOT–CNT are more likely than IrOx or TiN to permit effective stimulation of neurons. The search for a material with high sensitivity is critical for our goal of creating a high density MEA for stimulation of the LGN in blind patients. The next step for us in this project is to demonstrate that vision can be restored in a blind animals through electrical stimulation of the LGN in vivo.

Acknowledgment

All of the work reported here was funded by NPRP grant # NPRP 5-457-2-181 from the Qatar National Research Fund (a member of Qatar Foundation).

References

[1] Quigley HA and Broman AT. 2006. The number of people with glaucoma worldwide in 2010 and 2020. Br J Ophthal 90(3):262–267.

[2] Inayat S, Rountree CM, Troy JB, and Saggere L. 2015. Chemical stimulation of rat retinal neurons: feasibility of an epiretinal neurotransmitter–based prosthesis. J Neur Eng 12(1).