A study on electrical stimulation strategy for the encoding of visual information in neuronal activities of retinal ganglion cells

Abstract

Retinitis pigmentosa (RP) and age-related macular degeneration (ARMD) are retinal degenerative diseases leading to total blindness or severe visual impairment. Since there is no medical treatment for these diseases yet, researches on retinal implant have been actively conducted in order to artificially activate the remaining retinal ganglion cells (RGCs) by electrical stimulation and to transmit the visual information to the visual cortex. However, major efforts were focused on hardware such as electrodes and stimulators and relatively little attention has been paid to the pulse generation strategy for the retinal implant. As it is reported that effective stimulation strategy can improve speech recognition in cochlear implant, it is also expected that successful visual perception can be achieved with optimal stimulation strategies in retinal implant.Therefore, this dissertation aimed to develop an effective electrical stimulation strategy that can encode input visual information in neuronal activities of RGCs based on understandings of RGC response characteristics to electrical stimulation.In the first study (chapter 2), the author recorded neuronal activities of multiple RGCs simultaneously using multielectrode array (MEA) and classified RGCs into several types according to their temporal response patterns to the light stimulation. Since RGC responses vary according to the brightness of the light stimulus, it can be postulated that RGC responses encode input information (the brightness of the light stimulus). In order to check how much information is encoded in RGC responses, the author adopted a spike train decoding algorithm that has been mainly used in the brain-computer interface. By spike train decoding, input information was successfully reconstructed from RGC responses and results could be compared by quantitative values such as correlation coefficient between original and decoded results. From the results, it was found that more accurate decoding was possible with more RGCs and various types of RGCs since responses from same RGC types or synchronized-firing RGCs may contain redundant information. The overall result of chapter 2 suggests that the spike train decoding can be a proper approach to quantitatively evaluate how faithfully RGC responses encode input information. In the second study (chapter 3), the author investigated RGC response characteristics to electrical stimulation. First of all, it was found that RGC responses could be evoked and modulated by electrical stimulation within the certain current range. Under the assumption that temporal information such as light intensity variation (brightness) can be encoded in RGC responses by temporally-patterned electrical stimulation, pulse trains whose amplitudes were modulated according to various temporal patterns were delivered to the retina on the MEA. As a result, temporal pattern of electrically-evoked RGC responses accurately represent the pulse amplitude variation and input information delivered by electrical stimulation also successfully reconstructed from RGC responses by spike train decoding algorithms. From this result, the author suggested that electrical stimulation with amplitude-modulation can be a feasible mean to encode temporal information in RGC responses. The author also could check the possibility of the optimization of stimulation parameters (pulse amplitude range and pulse rate) by quantitative comparison of spike train decoding results. In the third study (chapter 4), the author investigated neuronal activities of RGCs of photoreceptor-degenerated retina, because the ultimate target of the retinal implant is a blind patient whose photoreceptor layer of the retina is degenerated due to ARMD or RP. Interestingly, though neuronal activities of the degenerated retina showed abnormal rhythmic patterns, RGC responses could be modulated by pulse amplitudes like those of normal retina. It was also confirmed that RGC responses of degenerated retina can be reliably modulated by not only pulse amplitude but also pulse frequency. In this chapter, two electrical pulse trains representing brightness of two pixels were applied to the retina simultaneously in order to check whether it is possible to encode spatiotemporal information in RGC activities by simultaneous multichannel stimulation. From the result, it was found that the effect of electrical stimulation was spatially limited on the MEA and RGCs encode information delivered by nearby stimulation electrode. Spike train decoding results supported the result since the information of two independent time-series delivered by two-channel electrical stimulation was accurately decoded from electrically-stimulated RGC responses. The results suggest that amplitude-modulation is an effective stimulation strategy to encode not only temporal information but also spatiotemporal information in RGC responses by electrical stimulation.