Abstract Neuroprosthetic retinal interfaces depend upon the ability to bypass the damaged photoreceptor layer and directly activate populations of retinal ganglion cells (RGCs). Current approaches to this task largely rely on electrode array implants. We are pursuing an alternative, light-based approach towards direct activation of the RGCs, by artificially causing them to express Channelrhodopsin II (ChR2), a light-gated cation channel. In addition to being non-contact, optical techniques lend themselves relatively easily to a variety of technologies for achieving patterned stimulation with high temporal and spatial resolution. In early studies, we are using viral vectors to obtain wide spread expression of ChR2 in rat retinas, and have developed a system capable of controlled large-scale, flexible stimulation of the retinal tissue with high temporal accuracy through adaptations of video projection technology. Finally, we demonstrate a pc-based wearable system that can perform the image processing transformations required for optical retinal neuroprosthetic interfaces in real time. I. INTRODUCTION OME of the more common causes of blindness are degenerative diseases of the outer retina, like Age- related Macular Degeneration (AMD) and Retinitis Pigmentosa (RP), globally afflicting approximately 25-30 million [1] and 1.5 million individuals [2], respectively. Diseases of the outer retina result in photoreceptor loss, while the inner retinal neurons and in particular the retinal ganglion cells (RGCs) and their optic nerve projections are largely maintained functional [3]-[5]. Artificial stimulation of these relatively well-preserved nerve cells with a retinal neuroprosthetic device could provide an artificial sense of vision by translating the visual scene into appropriate spatio- temporal patterns of neuron activity just as cochlear implants already translate environmental sounds into activity patterns of auditory nerve neurons. Current approaches towards artificial stimulation of retinal ganglion cells largely rely on the development of micro- fabricated electrode array implants which will either lie on the epi-retinal (over the retina) or sub-retinal surfaces [1]. While promising, this technology may suffer from several Manuscript received February 16, 2007. This work was supported by a Marie Curie International Reintegration grant #16596 and by the Kohns family fund. The authors are with the Faculty of Biomedical Engineering, Technion  Israel Institute of Technology, Haifa, Israel (phone: +972-4-8294125; fax: +972-4-8294599; e-mail: sshoham@bm.technion.ac.il). fundamental drawbacks including the long term stability of the interface between electrode arrays and the very fragile retina tissue and the challenge of creating interfaces with many thousands of independent channels in an attempt to approach highly functional vision. Alternative, non-contact, approaches could rely on techniques for direct light-based activation of the RGCs which can potentially cause reduced tissue damage. Combining such methods with existing spatial light modulation technology (micro-displays) can potentially allow a relatively simple route towards restoration of high-acuity vision. A system capable of patterned activation of multiple neurons with millisecond precision using rapid UV laser deflection and caged neurotransmitters, has recently been demonstrated by Shoham et al. [6] Another recent breakthrough in noninvasive control of neural activity is the discovery by Nagel et al. [7],[8] of Channelrhodopsin-2 (ChR2), a directly light-gated cation selective ion channel in the green algae Chlamydomonas reinhardtii. This membrane channel opens rapidly after absorption of a blue-light photon, generating a large permeability for monovalent and divalent cations. Thus, ChR2 is a novel tool which can be used to depolarize cells simply by illumination, and it has already been demonstrated that the neural activity can be optically controlled with millisecond precision in ChR2 - transfected hippocampal cell cultures [9]. Furthermore, Bi et al. [10] demonstrated that transfected mouse retinas with a recombinant adeno- associated virus (rAAV) can stably express ChR2-GFP channels in vivo, for periods of up to 1 year. They further demonstrated visual evoked potentials (VEPs) from ChR2- transfected blind rd1/rd1 transgenic mice subjected to full- field light flashes, suggesting that this technology may provide a viable path to vision restoration. II. METHODS A. ChR2 expression Viral vectors are currently the most efficient way to induce stable expression of genes in non-dividing CNS neurons, and studies suggest that adeno-associated viruses (AAV) are the most suitable candidates for prolonged expression of genes of interest in a retina [11],[12]. We injected the AAV2-CAG-Chop2/GFP-WPRE-BGH- polyA expression vector (GeneDetect Ltd.) introduced by Bi et al.[10] into the vitreous of rat eyes (a single eye per rat) Patterned optical activation of Channelrhodopsin II expressing retinal ganglion cells Inna Reutsky, David Ben-Shimol, Nairouz Farah, Shulamit Levenberg and Shy Shoham, Member, IEEE S Proceedings of the 3rd International IEEE EMBS Conference on Neural Engineering Kohala Coast, Hawaii, USA, May 2-5, 2007 ThC1.2 1-4244-0792-3/07/$20.00©2007 IEEE. 50