Visual Neuroprosthetics: Functional Vision for the Blind

Resent progress in materials and microfabrication technologies have allowed researchers to reconsider the prospect of providing a useful visual sense to the profoundly blind. This will be accomplished by electrically stimulating their visual systems via an array of implanted microelectrodes. The techniques of the semiconductor industry have been employed to create electrode arrays with three dimensional architectures. These arrays are proving to be safely implantible into the visual parts of the brain of animals with little significant long term consequences. Thus, the tools of neuroprosthetics have been developed to the point that they will soon be used to validate some of the physiological foundations upon which artificial vision have been based. Validation of these foundations will accelerate the rapid pace of this research. If these physiological underpinnings can be shown to be solid, a demonstration of functionally useful vision in blind human volunteers may be possible within a five year time frame.

Richard A. Normann Department of Bioengineering, and The John Moran Laboratories for Applied Vision and Neural Sciences The University of Utah Visual Neuroprosthetics- Functional Vision for the Blind Recent progress in materials and micro­fabrication technologies have allowed researchers to reconsider the prospect of providing a useful visual sense to the pro­foundly blind. This will be accomplished by electrically stimulating their visual sys­tems via an array of implanted microelec­trodes. The techniques of the semiconductor industry have been em­ployed to create electrode arrays with three dimensional architectures. These ar­rays are proving to be safely implantible into the visual parts of the brain of animals with little significant long term conse­quences. Thus, the tools of neuroprosthet­ics have been developed to the point that they will soon be used to validate some of the physiological foundations upon which artificial vision have been based. Valida­tion of these foundations will accelerate the rapid pace of this research. If these physiological underpinnings can be shown to be solid, a demonstration of functionally useful vision in blind human volunteers may be possible within a five year time frame. Overview In a research laboratory at the National Institutes of Health, a 42 year old female volunteer who has been totally blind for the past 22 years of her life "sees" very primitive patterns of light [1], At Johns Hopkins University, a blind man who has lost all visual sensation because of a pro­gressive retinal disorder, retinitis pigmen­tosa, sees for the first time in many years a small flash of light [2]. While these preliminary experiments are providing blind volunteers with only rudimentary visual sensations, this research is rekin­dling the hope that researchers may be able to provide the profoundly blind with a functionally useful visual sense. The research that may allow the reali­zation of this dream is still in its infancy. However, recent technological break­throughs in materials and in microfabrica­tion technologies afford vision scientists, neurosurgeons, ophthalmologists and en­gineers the opportunity to investigate this problem. This article reviews the physi­ological foundations behind visual neuro­prosthetics, and describes recently developed systems that may provide a use­ful visual sense to the profoundly blind. Figure 1 illustrates the basic compo­nents that a cortically based visual neuro- prosetic system will most likely contain. The system will require a video encoder to transform the visual world in front of a blind individual into electrical signals that can be used to excite neurons at some level of the visual pathway. This video encoder could be as simple as a miniaturized video camera mounted on a pair of eyeglasses. More likely, it will be a "silicon retina" that performs some of the image preproc­essing functions of the human retina and will make the signals compatible with the neurons they are intended to stimulate. The signals must be delivered to elec­trodes that will excite neurons at an appro­priate level in the visual pathway. This link will occur by either a hard-wired, percutaneous connection, or a telemetry link. Finally, the stimulating electrodes must be implanted into the visual pathway such that each electrode is able to excite only a small population of neurons in the vicinity of the electrode. Such a visual neuroprosthesis is expected to recreate a limited, but functionally useful visual sense in a blind individual who is using such a system. Physiological Foundations of Visual Neuroprosthetics An appreciation of the field of visual neuroprosthetics is predicated upon a ba­sic level of understanding of a number of well documented physiological observa­tions. These foundations are briefly de­scribed below. 1. Most forms of blindness are of retinal origin and leave the higher vis­ual centers unaffected. This observation is often unstated, but is key to a cortical approach to visual neuroprosthetics. The output neurons of the eye often degenerate in many retinal January/FebrtMiy 1995 IEEE ENGINEERING IN MEDICINE AND BIOLOGY 0739-517 5/9 5/$4.00© 199 5 77 1. A possible visual prosthetic system would consist of a video encoder to transform optical images into electrical images, and an electrode array implanted at some level in the visual pathways for their focal stimulation by electrical means. blindnesses. However, the neurons in the higher visual regions of the brain are usu­ally spared this degeneration. The higher visual centers of sighted individuals pro­vide a great deal of parallel information processing, and, in profoundly blind indi­viduals, this computationally rich, hierar­chically organized system lies fallow. If these higher visual centers can be stimu­lated with visual information in a format somewhat similar to the way they were stimulated prior to the development of total blindness, a blind individual may be able to use this complex neural computer to extract information about the physical world around him/her. 2. The visual pathways are organ­ized in a rational scheme. The organization of all sensory path­ways is characterized by "maps" that transform neural activity at one level in the nervous system into neural activity at other levels. Usually, the mapping is asso­ciated with some form of information processing. In the visual system dia­grammed in Fig. 2, there are a number of such maps. The light reflected from the objects in front of an observer (points "A" and "B" in this example) is projected by the optical elements of the eye onto an image on the back of the eye (the retina). This optical image is mapped into an elec­trical image by the photoreceptors of the retina. The cone photoreceptors decom­pose the colored image into three primary color maps (due to the absorption spectra of the three classes of cones) [3], The photoreceptor electrical image is mapped onto the retinal ganglion cells (the retinal output). Spatial, temporal, and chromatic filtering of the visual message take place during this mapping operation. Next, the retinal output (the optic nerves) projects to a relay station (the lateral geniculate nu­cleus) where binocular mapping begins. Another network of neurons (the optic radiations) map the image onto primary visual cortex located on the rear surface of the brain. There are many other maps in higher visual centers of the brain, but not a great deal is known of these maps [4], It is a general property of these map­ping processes that the maps of higher brain centers appear to be more complex than the maps of lower centers. Thus, the map representing the output of the cone photoreceptors would be expected to more closely resemble the pattern of light that strikes the photoreceptors than would the pattern of activity at the retinal ganglion cell level or the pattern of neural activity in primary visual cortex. In spite of this increased complexity of visual maps, it has been known for decades that the maps at the levels of the retinal output and the input to the visual cortex are reasonably rational (at least on the low resolution scale that they have been studied to date). A rational mapping of visual space onto the neurons of the visual cortex, while not yet fully demonstrated, is one of the fun­damental cornerstones upon which visual neuroprosthetics is based. 3. Electrical stimulation of neurons in the visual pathway evokes the per­ception of points of light (called phosphenes). This general physiological observation has also been known for decades, and is not particularly surprising once one appre­ciates the mapping concepts described above. Since the visual world is mapped onto a certain region of the cerebral cortex, artificially stimulating the neurons in this region would be expected to produce per­cepts similar to those that are evoked by natural excitation: i.e., points of light. Such artificial stimulation can be achieved by the injection of very small electrical currents via external wires. Recent experiments by researchers at the National Institute of Health [1] and others at Johns Hopkins University [2] have focused on obtaining a better under­standing of these phosphenes. Both groups have evoked phosphenes in human volunteers using electrical stimulation via a very fine insulated wire (an electrode), but the site of the stimulation differed. The Hopkins researchers placed their elec­trodes on the surface of the retina while the NIH team placed 38 electrodes into the visual cortex. Both groups demonstrated that small electrical current pulses (in the microampere range), evoked the percepts of spots of light in front of the blind vol­unteers. Both groups verified that on the coarse scale studied, the perceptual maps were as expected. These results suggest that passing a two dimensional set of elec­trical currents through an array of elec­trodes, inserted into the visual system at an appropriate location, should create the perception of an array of phosphenes. Since the apparant brightness of these phosphenes can be controlled by modulat­ing the currents through each electrode, this array could be used to produce a "pixelized" (or discretely represented) visual sense made up of a large number of adjacent phosphenes. The resulting visual experience might be something like that produced when looking at a football sta­dium scoreboard. An important corollary to these experi­ments has developed over the past decade: it is possible to pass these relatively low amplitude currents through microelec­trodes for long periods of time without appreciable damage to either the tissue into which the currents are injected, or the electrodes themselves [5]. 4. The visual system is a highly adap­tive neural network with a great capac­ity to make appropriate correlations between the sensed visual world and the physical world around the observer. 78 IEEE ENGINEERING IN MEDICINE AND BIOLOGY January/Februory 1995 Optic Radiation .. visual / Cortex S 2. Schematic view of the visual pathways. Adjacent points in space (A, B) are mapped into adjacent images on the retina. This map is transformed into an electri­cal map that is projected onto the visual cortex, which then excites groups of cortical neurons that are adjacent to each other. Lateral Geniculate Nucleus «. The plasticity of the visual system has been graphically demonstrated in experi­ments where observers viewed the world through inverting prisms for many months [6]. When first donning the prisms, the observers felt nauseous and had trouble maintaining balance. After a few days, they were able to move slowly in familiar environments and, after a few weeks, they were able to perform more complex motor tasks such as writing. In a few months, the subjects were able to ride bicycles and to catch thrown objects. This capability is only possible if the visual systems in these adult observers can be effectively rewired due to this completely abnormal visual stimulation. Thus, while the phosphene sense presented to a non-congenitally blind individual is expected to be a depar­ture from the individual's previous visual experiences, it is hoped that this tremen­dous neural plasticity will allow quick re­association of the "phosphene world" with the physical world. Previous Attempts at Artificial Vision The four observations described above provide a physiological framework around which a visual prosthetic system could be designed and built. As these ob­servations have been generally appreci­ated for decades, it is not surprising that a number of scientists and engineers have used the observations to make pioneering attempts at providing a useful visual sense for the blind. Two notable efforts at devel­oping an artificial vision system were un­dertaken in the 1960s and early 1970s. Giles Brindley and his collaborators at Cambridge University developed a proto­type visual prosthetic system and per­formed extensive tests with blind volunteers [7]. This system was based upon an electrode array that was placed upon the surface of the visual cortex. Each electrode was connected via its own lead wire to a small coil of wire. The coils of wire were mounted under the scalp over the skull of the blind volunteer, and al­lowed the passage of electrical current through each electrode via an inductive telemetry link. This telemetered approach minimized the likelihood of infection. William Dobelle and his colleagues performed most of their work at the Uni­versity of Utah and their findings were published somewhat later than Brindley's [8]. Dobelle also used an array of elec­trodes placed over the surface of the visual cortex, but, unlike Brindley, Dobelle used a percutaneous connector mounted behind the volunteer's ear to provide electrical access to the electrodes in his arrays. Dobelle performed a large number of acute experiments as well as a number of longer term semi-chronic experiments in blind volunteers. Both Brindley and Dobelle made simi­lar observations. Both groups demon­strated that individual phosphenes could be evoked by current injections through most of the electrodes in their arrays, and that the amount of current required to evoke a phosphene in each electrode was relatively constant during the entire period of the implantation. They both made the important observation that injection of currents through a number of electrodes could be used to produce complex patterns of phosphenes. However, neither en­deavor resulted in the development of a useful visual prosthesis. Dobelle's sub­jects were able to interpret the current injections through six selected electrodes as Braille characters. These subjects could read the phosphene generated Braille characters faster than they could read them with their finger tips. This is hardly a useful visual sense and would certainly not warrant the risks associated with sur­gical implantation of an electrode array. Moreover, Brindley's and Dobelle's experimental findings on minimal elec­trode spacing made the concept of a visual prosthesis based upon surface stimulation of visual cortex even less tenable. Because of the high resistivity of the membranes covering the surface of the brain (relative to the low resistivity of the underlying neural tissue), electrodes on the surface of the visual cortex would not be expected to produce focal excitation of the neurons underlying the electrodes. Because of this, the current levels required to evoke phosphenes were in the milliamp level, which precluded the possibility of simul­taneous stimulation of too many elec­trodes (seizures could have been produced). Further, because surface stimulation excited a large population of neurons under the cortical surface, the large surface electrodes could not be placed too close to neighboring elec­trodes. Phosphenes evoked by currents in­jected through pairs of electrodes too closely spaced would often interact in a highly nonlinear way. Artificial Vision in the 1990s Brindley's and Dobelle's work made it clear that functionally useful artificial vi­sion could not be possible until a means was found to provide much more focal stimulation of neurons in the visual path­way. Thus, much smaller stimulating elec­trodes than those used in the earlier studies were needed. These new generation elec­trodes would be about the same size as the neurons they are intended to stimulate. Kensall Wise at the University of Michi­gan applied the techniques used to build integrated circuits to the creation of small, high density electrode arrays [9]. This new generation of electrode array was built from silicon, and so it also contained inte­gral signal processing circuitry. A mi- Januory/February 1995 IEEE ENGINEERING IN MEDICINE AND BIOLOGY 79 3. Two silicon based microelectrode arrays with a two dimensional "comb" architec­ture. These electrode arrays are created using the IC photolithographic techniques. The arrays also contain integrated electronic circuitry. Photo courtesy of Kensall Wise, University of Michigan. ment of the brain. Any parts of the array that are to extend outside of the neural tissues must be as unobtrusive as possible. Finally, there are three-dimensional archi­tectural constraints that a cortical elec­trode array must satisfy. It is these considerations that have led researchers at the University of Utah to develop the cor­tical implant described below. The Utah Intracortical Electrode Array Rather than trying to adapt an existing technology to solve a particular problem, the Utah team first defined an electrode geometry that would be best suited to the task of cortical stimulation. Then they tried to fabricate such a geometry. The ideal array 1. must be able to stimulate a two di­mensional sheet of neurons that are copla- nar with the surface of the visual cortex. 4. Scanning electron micrograph of the Utah intracortical electrode array. This "hair brush," three-dimensional architecture contains 100 electrodes, each of which is 1.5 mm long, 0.08 mm at its base, and tapers to a sharpened tip. crograph of two of Wise's stimulating electrode arrays is shown in Fig. 3. Because the techniques used to create these arrays are basically the two dimen­sional photolithograpic processes used by the semiconductor industry, the array has a two dimensional geometry and can best be described as a "comb-like" architec­ture, where the teeth of the comb contain the active electrodes. The teeth are in­tended to be implanted into the cortical tissues. These new generation electrode arrays seem to have solved most the problems uncovered by earlier workers. The arrays are very small and the silicon from which they are fabricated is highly biocompat­ible. The arrays can be inserted into neural tissues, and the active electrode regions can be directly apposed to the neurons they are intended to stimulate to achieve very focal stimulation. However, there are other issues that must be addressed in the design of an electrode array that is in­tended to be used as the basis for a visual neuroprosthetic system. The integrated circuitry on the arrays must be hermeti­cally sealed from the corrosive environ- 2. must have electrodes that penetrate through the surface and down to the level of the normal neuronal input to the visual cortex: the optic radiations. In humans, this region is about 1.5 - 2.0 mm below the cortical surface. 3. must have array components that are as unobstrusive as possible (applies to any component that extends beyond the cortical surface). The electrode array architecture cre­ated by the Utah team is shown in Fig. 4. The techniques used in its manufacture has been described by Kelly Jones from our laboratory [10]. This array success­fully achieves the conditions described above. The three-dimensional architecute of this array can be described as a "hair brush' (as opposed to the comb metaphor of the Wise electrode geometry). The electrode geometry shown in Fig. 4 was designed for use in animal experi­ments. Each of the 100 needles in these arrays is 1.5 mm long and projects from a silicon substrate that can be as thin as 200 |j.m. The substrate is 4.2 mm on a side. The electrodes taper from an 80 |im base to a sharpened and metalized tip. Each elec­trode is electrically isolated from its neigh­boring electrodes by a glass dielectric that surrounds its base. Electrical access to each electrode is via a demultiplexing unit integrated onto the back surface of the array. However, as this circuitry has yet to be fully developed, we are currently mak­ing direct wire bond contact to a subset of 10 to 12 electrodes in the array, and con- 80 IEEE ENGINEERING IN MEDICINE AND BIOLOGY January/February 1995 5. Light evoked photoresponses recorded with the Utah electrode array. The re­sponses (top trace) were evoked by a vertical slit of light moving horizontally in front of a cat. The horizontal position of the slit is indicated in the middle trace. In the lower traces, three of the responses shown in the top have been superimposed and replotted on a faster time scale to demonstrate that these three large responses are probably from a single unit. necting these insulated lead wires to a percutaneous connector. The entire struc­ture, with the exception of the metalized tips, is insulated with 1 (im thick dielectric material. Insertion of Electrode Arrays Cortical tissue is relatively soft, but is covered with a tough membranous skin called the dura mater. Even when the dura mater is peeled back, it is a formidable task to implant a large number of electrodes into the cortex. While one or a few needles can be readily inserted, trying to insert a large number of electrodes only dimples the cortical surface. Thus, only partial in­sertion of electrodes around the perimeter of the array can be achieved. This problem has been solved with a pneumatically ac­tuated inserter system, as designed, built, and evaluated in our laboratory by Patrick Rousche [11]. The inserter is capable of accelerating the array to high velocities and "shooting" it into the cortical tissues. The viscoelastic properties of the cortical tissue are such that complete insertion of the arrays is possible with little tissue in­sult. Evaluation of Electrode Arrays The biocompatibility of the Utah elec­trode arrays has been evaluated in a series of acute and chronic histological and elec- trophysiological experiments. Electrode arrays have been chronically implanted in cats for up to fourteen months. Histology Our histological studies of cortical tis­sues that have undergone electrode array implantation are most encouraging [12]. To summarize our findings, each elec­trode becomes encapsulated, but normal appearing neuron cell bodies are seen 10 to 20 microns from each electrode track. A fraction of the electrode tracks indicate that a small amount of local bleeding had occurred around the track, probably as a result of the implantation procedure. There is evidence of edema in the vicinity of the tracks, but the level has been judged to be non-problematic. Electrophysiology Perhaps the most compelling evidence of the biocompatibility of the materials and the benign nature of the implantation process is that evoked neural activity can be recorded with the electrode arrays in response to light simulation in specifi re­gions of the animal's visual space. An example of the evoked activity is shown in Fig. 5. The response is made up of action potentials from a number of cortical neurons, typical of a unit cluster response. In one cat that had been implanted on a chronic basis, we were able to record simi­lar cluster responses (but manifesting a poorer signal-to-noise ratio than that shown in Fig. 5) for up to 14 months (at which time the animal was sacrificed). Craig Nordhausen, a graduate student in our laboratory, has conducted a complete study of the acute recording capabilities of the Utah intracortical electrode array [13]. Psychophysical Experiments The visual sense that stimulation of the visual cortex is expected to produce will differ in many respects from that of sighted individuals. Since pairs of adja­cent electrodes will not be able to be posi­tioned so close that current stimulation through both will produce two completely contiguous phosphenes, the visual sense is likely be somewhat pixelized (i.e., com­posed of a large number of very closely spaced phosphenes). Further, as first gen­eration arrays will be limited in the number of electrodes they contain, the scope of the visual sense evoked will be far from the panoramic one of sighted individuals, but will be much more of a "tunneled" nature. The size of this tun­neled sense will be limited by the radius of curvature and arc length of a cortical gyrus (bump on the brain surface). Based upon a number of anatomical and physi- January/Februory 1995 IEEE ENGINEERING IN MEDICINE AND BIOLOGY 81 6. Photograph of the portable, pixelized vision emulator, which is composed of a video camera and video monitor mounted on a pair of ski goggles. The monitor is masked with perforated foil, which simulates the pixelized visual sense that the blind subject might experience when stimulated via the Utah intracortical electrode array. ological criteria [14], it is unlikely that electrodes could be usefully spaced closer than 300-400 |im. Thus, the visual sense created with a first generation visual pros­thesis might consist of about 625 phosphenes distributed over a 1.7 degree visual angle (about the size of one's thumb nail at arms length). When confronted with these discouraging quantitative con­siderations, one wonders if such a limited and pixelized visual sense could be of any use to a blind individual. Thus, before electrical stimulation of the visual cortex could be considered to be a viable ap­proach, experiments would be needed to evaluate what kind of visual task perform­ance can be achieved with a very limited pixelized visual sense. Using a portable pixelized vision emu­lator (Fig. 6), Kenneth Horch and Kichul Cha have recently performed a number of experiments to address this question. The emulator is described in [15], and consists of a miniaturized, closed circuit video sys­tem mounted on a pair of ski goggles. The only visual input to the wearer of the emu­lator is via a miniaturized video monitor covered with a perforated foil mask con­taining up to 1024 holes. Optics in front of the monitor limit the size of the perceived pixelized image to 1.7 degrees of visual angle and a lens in front of the video camera controls its acceptance angle. Scanning of objects in front of the wearer is accomplished by head movements. Us­ing this device with a foil mask containing 625 perforations, student volunteers are able to read text out loud from a computer screen (at a rate about 2/3 that at which they can read without the emulator) [16]. Much of the work we have done with the simulator has been focused on determin­ing the degree of visually guided mobility that can be achieved with such a limited pixelized visual sense. We have built a programmable maze through which stu­dent volunteers must navigate, while us­ing only this limited pixelized visual input. The maze is formed by a large number of curtains that can be raised or lowered by the experimenter. After each traversal of the maze, curtains are adjusted and a new path created. The index of visually guided mobility is the time required to navigate the maze. After about 30 trips through the maze, the learning curves of the student volunteers flatten. With only 625 pixels of visual information, the subjects were able to navigate through this highly complex visual environment at a speed close to the speeds attained with normal vision. More importantly, they navigate this complex environment with high levels of confi­dence, as described elsewhere in [17]. Discussion While the focus of this article was elec­trical stimulation of the visual cortex, a thorough exposition of artificial vision must also consider work directed at elec­trical stimulation of the ganglion cells of the retina. The concept being pursued by researchers at Johns Hopkins University [2] and at MIT and Harvard University [18] is to implant a visual prosthesis di­rectly on the surface of the retina of a blind indvidual. The implant is proposed to con­tain an array of photodetectors on its front surface and an array of electrodes on its rear surface. Signal processing electronics will also be incorporated into the implant. In support of this concept, the Hopkins researchers have stimulated, with a single microelectrode, the retinas of a number of patients who have advanced retinitis pig­mentosa. As expected, electrical stimula­tion of the retinal ganglion cells evoked phosphenes in these volunteers [2]. This retinal work is also in its infancy but it has a number of aspects that are particularly attractive. The coupling of electrodes and photosensors in a single implant could take advantage of the image forming properties of the eye (i.e., the cornea and lens) to transduce visual im­ages incident upon the photodetector array into electrical signals. This process could eliminate the need for an external video camera. The surgical implantation of a retinal device might be regarded as being less traumatic than a cortical implant. Since the retinal implant is closer to the visual image than is a cortical implant, the retinal map would be expected to be more rational than a cortical map. A retinal implant also has a number of considerations that would tend to argue against this approach. Most importantly, most retinal degenerations produce retinal ganglion cell degeneration. If this is the case, a retinal implant might have applica­bility in only a very small number of pro­foundly blind individuals. Further, the retina is an extremely delicate tissue (0.5 82 IEEE ENGINEERING IN MEDICINE AND BIOLOGY January/February 1995 mm thick), which is likely to react cata­strophically if a chronic implant of any significant size and mass were to be placed against and attached to it. If the extraocu­lar eye muscles are not immobilized, the saccadic motions of the eye could be large enough to produce trauma to the retina where the implant is attached. Finally, the retinal ganglion cell map is only rational if one is able to stimulate the retinal gan­glion cell bodies. As the optic nerve fibers that emanate from the retinal ganglion cell bodies and exit the retina at the optic nerve head are located between those cell bod­ies and the surface electrodes, a technique by which the retinal ganglion cell bodies can be stimulated with currents that do not stimulate the optic nerve fibers will have to be developed. This last problem is likely to present quite an interesting challenge. It is clear from this discussion that demonstrating functional artificial vision in a blind volunteer is still many years away. While some progress has been made in terms of phosphene charac­terization in blind human volunteers, elec­trode array development, implanting such complex structures in the brain, and in the biocompatibility of these cortical im­plants, there are many issues that must be resolved before chronic human implanta­tion of high density electrode arrays can be considered. Animal behavioral experi­ments should provide us with a better un­derstanding of the problem of electrode interactions. Once groups of animals have been trained to press levers when they perceive a point of light, we can inject electrical currents into their cortices rather than present them with a visual stimulus. If the animal presses the lever, the electri­cal stimulation is presumably evoking a percept similar to that produced by visual stimulation. These experiments will allow us to measure phosphene thresholds. Similar current injection with animals trained to respond to two spots of light will provide better criteria for spacing of elec­trodes in cortical arrays. With the development of these new three-dimensional electrode arrays, we will soon be in a position to explore the fundamental physiological foundations of artificial vision. Specifically, just how precise is the mapping of visual space onto the visual cortex? How stable are the phosphene thresholds on a day by day basis, even when the visual cortex is being stimulated as much as 16 hours per day? How far apart can a pair of electrodes be positioned and still produced contiguous phosphenes. Does patterned stimulation (passing a set of currents through a prese­lected set of electrodes) produce patterned percepts? When these fundamental questions have been answered, the intellectual bar­riers to a visual prosthesis based upon electrical stimulation of visual centers will be removed, and the question of providing useful vision to the blind will be "when" rather than "if'. An optimistic view of the pace of research in artificial vision must be based upon the results of these physi­ological experiments. It is not unreason­able to expect that an electrode array that could provide a useful visual sense to a blind volunteer could be implanted within the next five years., and commercial visual prosthetic systems could be available af­ter the turn of the century. Acknowledgments The work described in this manuscript was supported by grants from the N.S.F., the W.M. Keck Foundation, the John A. Moran Trust, and the Texaco Research Foundation. Richard A. Normann is Professor and Chair of the Department of Bio­engineering at the Uni­versity of Utah, where he has conducted basic research on information processing in the verte­brate retina and visual cortex and applied research in the area of visual neuroprosthetics. The Bioengineer­ing Department has recently been awarded a Whitaker Development Grant to create an academic program in "Bio­based Engineering." The goal of this pro­gram is to bring some of the organizational principals of biological systems into the discipline of Bioengineering and to apply some of the problem solving strategies of biological systems to the problems of con­ventional engineering and medicine. Pro­fessor Normann can be reached at the Department of Bioengineering, 2480 MEB, The University of Utah, Salt Lake City. UT 84112. References 1. Vallabhanath P, Bak MJ, Hambrecht FT, Kufta CV, O'Rourke D, Schmidt EM: Feasibil­ity of an intracortical visual prosthesis for the blind: II. Visual sensations produced by cortical stimulation. Soc. Neurosci. Abstr., 18, Part 2, 1396, 1992. 2. Humayun MS, Propst RH, Hickingbotham D, dejuan Jr. E, Dagnelie G: Visual sensations produced by electrical stimulation of the retinal surface in patients with end stage retinitis pigmen­tosa (RP). Assoc. Res. Vis Ophthal. Abstracts 835, 1993. 3. Normann RA, Perlman I, Hallett PE: Cone photoreceptor physiology and cone contributions to color vision. In: Gouras P (ed): The Perception of Color. In: Vision and Visual Dysfunction. Vol­ume 6. MacMillan Press, 1991. 4. Hubei D: Eye, Brain, and Vision, Scientific American Library, 1988. 5. Agnew WF, McCreery DB (eds): Neural Prostheses, Prentice Hall, 1990. 6. Dolezal H: Living In A World Transformed, Academic Press, 1982. 7. Brindley GS, Lewin WS: The sensations pro­duced by electrical stimulation of the visual cor­tex, ,/. Physiol 196: 479-493, 1968. 8. Dobelle WH, Mladejovsky MG: Phosphenes produced by electrical stimulation of human oc­cipital cortex, and their application to the develop­ment of a prosthesis for the blind. J. Physiol 243: 553-576, 1974. 9. Wise KD, Najafi K, Ji J, Hetke JF, Tanghe SJ, et al: Micromachined silicon microprobes for CNS recording and stimulation. 12th Interna­tional Conference of the IEEE Engineering in Medicine and Biology Society 5: 2334-2335, 1990. 10. Jones KE, Campbell PK, Normann RA: A glass/silicon composite penetrating intracortical electrode array. Annals of Biomedical Engineer­ing 20: 423-437, 1992. 11. Rousche PK, Normann RA: A method for pneumatically inserting an array of penetrating electrodes into cortical tissue. Annals of Biomedi­cal Engineering 20: 413-422, 1992. 12. Schmidt S, Horch KW, Normann RA: Bio- compatability of silicon based electrode arrays implanted in feline cortical tissue. J. Biomedical Materials Research 27, 1393-1399, 1993. 13. Nordhausen CT, Rousche PJ, Normann RA: Optimizing recording capabilities of the Utah Intracortical Electrode Array. Brain Res, 637, 27­36, 1994. 14. Normann RA: A penetrating, cortical elec­trode array: design considerations. 1990 IEEE International Conference on Systems, Man, and Cybernetics 918-920, 1990. 15. Cha K, Horch KW, Normann RA: Simula­tion of a phosphene-based visual field: visual ac- cuity in a pixelized vision system. Annals of Biomedical Engineering 20, 439-449, 1992. 16. Cha K, Horch KW, Normann RA, Boman DK: Reading speed with a pixelized vision sys­tem. J. of the Optical Society of America. 9:673­677, 1992. 17. Cha K, Horch KW, Normann RA: Mobility performance with a pixelized vision system. Vi­sion Research 32:1367-1372, 1992. 18. Risso, J: Development of a silicon retinal implant. Assoc. Res. Vis Ophthal. Abstracts, 1994. January/February 1995 IEEE ENGINEERING IN MEDICINE AND BIOLOGY 83