Michael L. Leonhardt**
David A. Hodges***
Elwin Marg****
*University of California at Berkeley
**Current address:IBM Corporation San Jose, Ca
.***Department of Electrical Engineering and Computer ScienceUniversity of California Berkeley, California
****School of OptometryUniversity of California Berkeley, California
Reprinted from: The Proceedings of the San Diego Biomedical Symposium, Jan. 31 - Feb. 2, 1973 12: 247-259.
ABSTRACT
A phosphene visual prosthesis is described that utilizes 1024 electrodes on the visual cortex of the human brain. Information to energize the electrodes, derived from a photosensor array, is transmitted by digital time division multiplex technique along with system power over radio channels through the scalp. This battery powered system is compact and portable with most of the internal electronics mounted on the back of two flexible electrode arrays. External components are housed in a single, small package.
INTRODUCTION
Although there are many kinds of sensory aids for the blind, most of them make direct or indirect use of auditory or tactile information which is then interpreted by the subject to give spatial information. All of these systems have a limited information flow and require a good deal of learning before they are optimally developed. The ideal goal would be a system to restore sight to the blind.
Almost all blindness is caused by some disease or destruction in the eye or optic nerve, which are actually peripheral parts of the visual system. There is the possibility of substituting an optical electronic device to replace these damaged parts of the visual system and provide a direct visual experience in blind individuals. Such a device would require little or no learning for its use and have a higher flow of information than other types of visual prostheses. The state of the electronic art today includes the capability of building such devices. The problem, once they are built, is how to couple them effectively and safely to the visual system.
Stimulation of the optic nerve, which is accessible in the orbit as it comes from the eyeball is one possibility. However, microelectrodes to stimulate micronerve fibers have not been successfully developed as yet. In addition, the introduction of electrodes in the optic nerve fiber bundle would cause destruction of many of the nerves, leading to the death of these visual cells. Infection is a potential problem since the electrode site would not be well sealed from the outside environment. Finally, many common blinding diseases, such as glaucoma, destroy the optic nerve making coupling to it useless. Similar objections exist for stimulating the visual pathway as it courses through the brain. The lateral geniculate body is a prominent, well-known visual station on the way from the eye to the visual cortex that is accessible by stereotaxic brain surgery. However, no successful method for stimulating on the microscale of the nerve fibers or cells in this nucleus has been developed.
The surface of the visual cortex itself seems the best choice for a coupling junction. The visual cortex is divided into three separate areas on the basis of cellular architectonics, areas 17 (also called the striate cortex), 18 and 19. A percept of light is elicited when the visual cortex is stimulated electrically and this percept is termed an electrical phosphene. Phosphenes are commonly observed when the brain is stimulated by other effective stimuli such as a blow on the head. Then it is popularly termed "seeing stars."
Foerster (1929)1 was the first to demonstrate a phosphene from electrical stimulation of the visual cortex under controlled conditions. He reported that the phosphene was stationary and punctate. Krause and Schum (1931)2 confirmed this finding and demonstrated that stimulation of the visual cortex could still elicit phosphenes even though the visual input to it had been absent for a long period. It had previously been thought that transneural degeneration might cause destruction of the visual cortex similar to that observed in the fibers of the lateral geniculate body leading to the cortex when visual inputs cease from degeneration of theoptic nerve.
The first case of cortical stimulation for a crude visual prosthesis was reported by Button in 19583 and later by Button and Putnam in 19624. Stainless steel wires were implanted through the scalp and into the visual cortex. Their uninsulated tips served as electrodes. Driven by signals derived from a photocell, this device allowed two of three blind subjects to locate a light source by manually scanning the photocell until the resulting phosphenes were brightest.
Two important features were lacking from this device for it to be considered a prototype for a useful prosthesis. The connections through the skin provide potential channels for infection. Further, it is physically impracticable to connect enough of these electrodes to produce an image with satisfactory resolution; a useful phosphene image would require 500 or more electrode points. Transmitting information and power through the intact skin would solve the first of these problems.
A design for such a transmission system using radio frequencies was published by Brindley in 1965, 5,6 and was incorporated in an implanted visual prosthetic device by Brindley and Lewin in 1968 7. Although many of the 80 electrodes in this prosthesis were not functional, it demonstrated that the approach was basically valid and that multiple, individually-controlled phosphenes could be elicited to provide a visual image. At the time of this writing, some five years later, the device is still in place and producing about ten different phosphenes during periodic tests. Those channels which have dropped out have done so suddenly without any systematic increase in the required electrical stimulation to produce phosphenes (i.e., the phosphene threshold); of the ten channels remaining, there has been no trend of a change in threshold 8. These observations indicated that the loss of channels is attributable to electrical causes rather than problems of the electrode-tissue interface. Since this first implant, papers have been published by Brindley and others 9,10,11,12. These reports include Brindley's second implantable system utilizing 180 electrodes with a similar type of radio frequency transmission as in the first implant.
A smaller device with 75 electrodes was actually implanted, with each of the 67 channels which functioned eliciting a phosphene 13,14. These results along with those from the first implant and preliminary work done with baboons 7 further support the practicability of a phosphene visual prosthesis and indicate that electrical stimulation of the cortex can be safely done.
In order to take full advantage of the visual field, as many phosphenes as possible must be placed in it. Brindley has evidence that stable phosphenes can be elicited from visual areas beyond area 17 13,15. On this basis it is not unreasonable to conclude that the whole visual cortex surface is functionally available for electrodes.
Although the implanted systems of Brindley and his colleagues are functionally sound, the system's electronics are not sufficiently sophisticated to allow them to be "scaled up" to 500-1000 electrodes, primarily because of space limitations within the head. With Brindley and Lewin's physiological data as a starting point, making use of multiplexing to reduce the number of external to internal data channels ant microelectronics to reduce circuitry size, as first proposed by Marg et al. 16 in a 512 electrode system, a 1024 electrode prosthesis can be realized using today's MOS large scale integration and bipolar technologies.
ORGANIZATION OF THE SYSTEM
The system is limited by physiological and electronic constraints. Of the physiological requirements, the possible noxious effects of electrical stimulation and the number of individually resolvable phosphenes are of primary concern. Once established, these requirements define the nature of the electronic hardware which is itself limited by speed, power requirements and packaging considerations. It is assumed that no physiological limitation will arise in the rate of information transfer through the parallel channels.
PHYSIOLOGICAL REQUIREMENTS
STIMULATION
Reliable production of phosphenes depends largely upon the stimulation duration and amplitude. Generally, the longer the duration the smaller the required amplitude but the greater possibility of irreversible activity at the electrode tissue interface. Typical thresholds for stimulation durations of 200 usec are on the order of 10 volts. Pulse stimulation is the most effective but there should be no net charge flow into or out of each electrode on the cortex. Any such charge transfer can produce toxic electrolysis and may cause lesions. This requirement could be met by using bipolar symmetrical pulses or capacitive coupling between each electrode and its stimulation source. However, the space taken up by presently available types of discrete capacitors in a system with a large number of electrodes would be prohibitive. Investigation is being done by Guyton 17 on a capacitive-type tantalum electrode using a dielectric at the surface of the electrode tissue interface, but the present devices may have too low a charge transfer to be useful. Such electrodes would allow the use of both types of net charge flow protection. Using a stimulation pulse duration in the 200-400 usec range may be most desirable since the lower phosphene threshold voltages of longer duration pulses are retained but the period that charge flows in any one direction during stimulation is minimized.
The stimulation repetition frequency has only a minor effect on phosphene thresholds but frequencies in the 50-100 Hz range have been found empirically to be the most effective. All other things being equal, a lower frequency results in lower operation speeds, and reduced power consumption for the electronic circuits and a smaller duty cycle for any one electrode, allowing more electrodes to be multiplexed to common stimulating circuitry. Since these are all desirable features, 50Hz is the preferred stimulus frequency.
The amount of voltage necessary to produce a threshold phosphene varies from electrode to electrode on the cortex. There is a limited range from the value of this threshold voltage or current to the point where increased phosphene persistence would blur the image and perhaps damage the brain (about 1.5 to 2.5 times the threshold voltage) over which varying stimulation pulse amplitude will affect the phosphene brightness. Therefore, we can incorporate a gray scale into the system design. Because of the limit of the dynamic stimulation range, 32 levels of intensity should be more than sufficient for a practical gray scale. These levels could be either logarithmic or linearly graded steps. It is not presently known which is more effective. Figure 1 summarizes these stimulation requirements.
THRESHOLD MAPPING
To have a uniform gray scale for all the electrodes we must take into account the varying threshold voltages. These variations would be mapped experimentally after the device was implanted and correction factors for each electrode stored digitally in a read only memory (ROM) in the external electrode package. To have a sufficiently high voltage for all thresholds plus gray scaling, at 200-400 usec pulse widths, the stimulation electrodes require a maximum of 15v. The stimulation is bipolar-symmetric; therefore we need +-15v internally to power the electrodes.
POSITION MAPPING
There is no well defined mapping between the array of electrodes on the visual cortex and the array of phosphenes produced on the visual field. The point to point mapping from cortex to phosphene must be found experimentally and the information stored in a second external ROM so that each phosphene produced will be supplied the correct intensity value for its point in the overall image. The computerized mapping scheme of Mao, Hodges, and Marg could be adapted to this device 18.
ELECTRODES
The size of the surface area on the visual cortex which can produce phosphenes when stimulated places an upper limit on the number of electrodes possible. A minimum electrode is approximately 1mm^2 for reasons of impedance and current density. The electrode material would be a physiologically inert substance such as platinum. Taking into account all three visual areas, 17,18 and 19, there is space (approximately 40cm^2) for 100 to 1000 such electrodes on each hemisphere depending on interelectrode spacing. As was noted in the introduction, 500 electrodes are probably about the lower limit for a useful prosthesis. A 1024 electrode system is chosen because it represents a conservative electrode spacing with a better chance of phosphene resolution and is equal to 21° for ease of digital processing organization. This will be discussed in more detail later. Separate 512 electrode arrays will be placed against the visual cortex of each hemisphere.
EXTERNAL TO INTERNAL CONNECTIONS
Incisions through the scalp and skull cannot be used to pass connections between external and internal circuitry because of the danger of infection through the opening. All external to internal connections are made by inductive coupling.
INTERNAL POWER
Power necessary to operate stimulating electrodes and internal circuitry must be inductively coupled from without to within 19. Internal power requirements should be kept to a minimum. Internal batteries are not acceptable as they would require subsequent surgical operations for replacement and are too bulky for implantation. A power oscillator circuit must be provided in the external electronics. The stimulating electrodes account for the majority of internal power consumption. In designing an electrode multiplexing scheme the preferable arrangement would be one where a minimum number of electrodes is stimulated simultaneously.
PACKAGING AND MATERIALS
Any device to be put in the internal electronics package must be packaged such that it is nontoxic to the body and unaffected by a 40° C saline environment. Integrated circuitry must be sealed to protect against sodium contamination. A vacuum deposited layer of silicon nitride over the chip would provide this protection; mechanical protection would be provided by layers of medical silastic 14.
Secondary coils of the inductive links are platinum, spirally deposited on a single, thin, flexible plastic substrate. This coil assembly would be placed between the scalp and skull and have wires leading from it through a burr hole in the skull to the circuitry and electrodes on the visual cortex.
HARDWARE
GENERAL SYSTEM CONSIDERATIONS
The electronics is divided between internal and external units. A minimum of the system should be internal to reduce the implanted package size, power consumption, power dissipation, and to maximize reliability.
Because of the large number of electrodes used in each group (512), it is impractical to run wires between the electrodes and an electronics module. For this reason, the electronics is mounted on the backside of the stimulating electrode substrates. There will be two internal packages each consisting of a 512 electrode array and its associated electronics which will rest against the visual cortex. The only parts not in these packages themselves are the secondary coils of the inductive links and any associated buffering circuits.
Inductive link coils are closely-coupled to minimize the possible effects of outside interference 19. The primary coils require some means of being held on the head above their respective secondary coils -- a special hat, wig, or headband could serve this function; the fewer the number of coils, the simpler this piece of headgear would be. A trade off exists between the amount of circuitry used and the number of inductive links. Information can be transferred between external and internal electronics in parallel or serially. Parallel techniques require more inductive links; serially require more internal circuitry. Since Large Scale Integration (LSI) circuits chips are used in the electronics, the cost of additional internal circuits is small so the number of inductive links is minimized.
Because the primary and secondary coils in the links may not always be in alignment, the system must be tolerant of losses occurring in these links. Attenuation of electrode stimulus signals would throw off the externally programmed threshold voltage and effect gray scaling. Using digital signals in the inductive links greatly increases the system's tolerance for misalignment of the coupling coils and provides additional immunity from outside interference, but introduces the additional complexity of internal digital to analog converters.
Picture information could be gathered by a television camera but the system does not require that much resolution. For the sake of compactness and simplicity a more practical source of images is a matrix of phototransistors. An image matrix for this system (of 1024 points) could be realized by a 32x32 array on a single integrated circuit chip whose picture elements are accessed by a 10 bit digital address, Such devices are available commercially 20 and packaged with a lens could be conveniently mounted on an eyeglass frame.
Because digital signals are required in the
inductive links, and the mapping information for electrode
threshold voltage and positions is stored digitally in the ROM's,
it is most convenient to handle the picture signals digitally.
Analog intensity signals from the image matrix are amplified then
passed through an analog to digital (A-D) converter for
digitizing. Most likely, some analog image processing steps would
be accomplished along with theamplification step.
Differentiation, to enhance the contrast of the picture
information, and an automatic gain control would be helpful.
However, only experimentation can determine what specific types
of image processing are most useful for cortical stimulation.
Further discussion of optimal types of picture processing is
outside the range of this paper. As a starting point the
prototype system would use straight amplification of the picture
signals before passing them to the A-D converter. Figure 2 shows
the functional structure that the preceding discussion
establishes for the system. 
ELECTRODE MULTIPLEXING
Electrode multiplexing enables us to minimize the number of inductive data links and the system electronics. Each electrode has a memory cell/driver circuit that can store a particular stimulus level then drive its electrode at that level until a new level is loaded into the cell. Given the requirement for say 400 usec to 1 msec bipolar symmetric stimulation, a memory cell would be accessed and loaded to a stimulus amplitude, be allowed to drive its electrodeat that amplitude for 200 usec then be reaccessed and loaded to the negative of that original amplitude and let run for another 200 usec. Finally, the cell is accessed a third time and loaded to zero, completing the stimulus cycle (see Figure 3).
Applying this cycle to a group of 8 electrodes would take 600 usec when a cell is accessed and loaded every 25 usec. Four, 8 electrode groups are run through this 600 usec cycle simultaneously so 32 electrodes are energized. The total system has thirty-two, 32 electrode groups, 16 per 512 electrode module. In practice the system nominally delivers an overall electrode repetition rate of 50 Hz. This defines a bipolar pulse width of
1 / 50Hz = 20 msec/cycle
20 msec / 32 electrode groups = ~625 usec/electron stimulus cycle
2*625 usec / 3 = ~417 usec bipolar pulse width
However, the system can be run up to a 100 Hz rate by varying the system clock frequency (adjustable externally) in which case the bipolar pulse width is approximately 209 usec. As is shown in Figure 4, in 417 usec 4x8=32 electrodes are energized.
Given the requirements for bipolar symmetric stimulation, the means must be provided to load the required electrode level into a storage element (which in turn sets the electrode's energizing amplitude), come back in 100-200 usec to unload that voltage level and load in its corresponding negative value, and after another 100200 usec set the stimulus value at zero. In this way a group of memory cells can be quickly loaded to the proper electrode amplitude values and run the electrodes for a 100-200 usec period while many other memory cells are being loaded or unloaded.
This electrode stimulation scheme is illustrated in Figure 3. During the first 200-400 usec as many electrode memory/drive cells are energized as time allows. Upon loading, these cells immediately set their electrodes to a level determined by the stored cell voltage. At the end of this 200-400 usec period, it should be recalled that there are 512 electrodes on each internal package. The two packages share common inductive data links whose use is alternated between packages. Thirty-two electrodes of one package are energized in typical 625 usec cycle, then 32 electrodes of the other package are energized. Next, the second 32 electrodes of the first package, and then the second 32 electrode group of the second package. This-continues until all 16 groups in each package have been energized once. All 1024 electrodes have now been energized in 32 consecutive 625 usec cycles or 20 msec -- this verifies the repetition rate of
1 sec / 20 msec = 50 stimulations per second
ELECTRODE MEMORY CELL/DRIVER
Figure 5 illustrates the memory/driver cell used for each electrode. When an electrode's address is selected, Q1 is turned on and output of the D-A converter is stored on capacitance C which is the intrinsic gate capacitance of Q2 and Q3. Depending on the polarity of this stored charge either Q2 or Q3 is turned on supplying the electrode either a positive or negative voltage at an amplitude determined by the voltage on the gate capacitance, C. To reload, Q1 is energized again changing C's stored voltage level to the present D-A output voltage.
ELECTRONICS PACKAGES
A rough overall system structure has been defined from physiological requirements and hardware considerations. An electrode multiplexing scheme and memory/driver cell circuit has been presented. The detailed structures of the electronics packages making up the system are now considered.
INTERNAL PACKAGE
Figure 6 depicts the structure and Figure 7 the timing diagram of one of the two 512 electrode internal electronics packages.
Five inductive links hook to the package to supply "store", clock and, reset pulses, intensity data and power. When the power is initially turned on, a pulse on the reset line zeros all the counters and latches. Intensity data is fed into a 23 bit shift register serially via the intensity line. The clock signal shifts the registers until they are filled with four 5 bit intensity data words. The bit between word 3 and 4 goes to all D-A converters to determine their polarity (for producing the bipolar signal). Extra bits between words 1-2 and 2-3 are not used.
A pulse on the store line enables the latches which in turn apply the intensity words into 4 D-A converters. It should be recalled that the D-A converters have 32 levels of output, either positive or negative as determined by the control bit. The "store" pulse also increments the 8 counter whose 3 tit output is fed to a 1 out of 128 decoder along with the 4 bit output of the electrode group counter (16 counter). Since the 16 counter has been reset when power was applied, it is in its initial state selecting addresses 1-8. The decoder does not select one of the 128 possible lines until it gets an enable signal. The purpose of using this enable signal is to inject a pause to give the power switch enough time to switch "on" or "off" if power is just being shifted to or from the other electronics package.
This power switch is the method by which the external package selects which internal package will handle the intensity information being sent. Both internal packages always receive power to their counters, registers and latches but the D-A and memory/driver cell power is controlled by the power switches.
The first four intensity words that were shifted into the registers have been put in the latches so now four new intensity words are started into the shift register. At the 5th clock pulse the decoder is enabled and the four memory cells hooked to the first address line are selected and receive intensity levels from their respective D-A converters. The decoder is kept enabled long enough for the gate capacitances in the memory cells to charge. At the end of this period the enable line goes low; another clock pulse later, the new intensity information is shifted into the latches.
This cycle will repeat itself 8 times, then the external electronics sets the D-A polarity pulse high (negative polarity) and the same four sets of 8 electrodes are accessed again. For the third cycle the intensity information is all zeros and the 32 electrodes are reset to zero volts.
Now there have been three cycles of the 8 counter. The divide by 3 outputs a 1 that simultaneously sets the NAND gate supplying the power switch high shutting off the power in this package, sets the same gate in the other package low turning on power in the other internal package and increments the 16 counter.
The second internal package will go through the same three- cycles as this first one then power will be transferred back "on" in the first package. Three more cycles of 32 electrodes will now commence, but now the 16 counter is in its second state which is outputed to the decoder so addresses 9-16 are selected. Finishing address 16, power is again shifted and the second package cycles through its four sets of 9-16 addressed electrodes. Power will be shifted hack and forth 16 times each time incrementing the cycle counter. Referring back to Figure 4, note that these sixteen cycles accomplish the energizing of all the 1024 electrodes. Continuing in time the complete stimulation cycle starts to repeat itself.
COMPONENTS
Circuits in this internal package would use either complementary MOS or dynamic MOS technology circuit chips with beam leads. The one exception is the power switch; it would be a multiplexing gate type switch for speed and low power loss, using Schottky diodes and clamped bipolar transistors.
Reliability is a key factor here as parts replacement in the internal packages is virtually impossible. Redundancy of components is probably the only reasonable approach to try to increase reliability, but space is limited on the electrode substrate where these beam-lead chips are mounted. Beam-lead chips are limited to about 32 connections coming off each. This means the decoder and memory/ driver cells would take up 20 chips with other circuitry accounting for 7 chips. Therefore, shift registers, latches and counters could be duplicated, but the decoder and memory/driver cells could not. All circuit chips used would have to be carefully selected and tested.
EXTERNAL PACKAGE
The external package supplies the control signals, digital intensity signals and power for the internal packages. Figure 8 shows the package structure; Figures 9 and 10 are the timing diagrams for its internal control signals.
As can be seen from the diagram, the external package's configuration has been determined in large part by the directions taken in the internal package design. Four pairs of ROM's contain the threshold voltage and cortex to visual field mapping characteristics of all 1024 electrodes. When the system is cycling, the internal packages are supplied with information via the ROM that corresponds to the particular electrodes being energized.
MODES OF OPERATION
There are three modes of operation through which the system cycles. These modes correspond to the three cycles discussed for the internal package.
Mode 1: corresponds to the internal packages loading memorycells positive. Externally, the picture information goesdirectly from the image matrix through analog processing and A-D conversion to the intensity inductive link driver and simultaneously is written into a random access memory (RAM).
Mode 2: corresponds to the internal packages unloading, and then reloading memory cells negative. The ROM's are no longer enabled. Picture information is read out of the RAM and into the intensity link. The D-A polarity bit is set to 1 producing negative polarity D-A outputs in the internal packages.
Mode 3: corresponds to the internal packages unloading to zero. RAM and ROM's are not enabled. The multiplexer's latch receives a control signal setting all outputs low so a string of zeros is sent to the intensity link.
Each time a 3 mode cycle is completed (32 electrodes are energized during this cycle) the 8 counter is incremented. When this counter reaches a count of 8 (after 8, 32 electrode cycles i.e., 256 electrodes have been energized) it outputs a pulse to the "chip selector" which in turn selects the next set of ROM's (remember each set of ROM's only holds information for 256 electrodes).
TYPICAL OPERATION
When the power is initially turned on, a rest pulse is generated that resets all internal and external counters, latches, and registers. A start up sequence must then be initiated to get everything synchronized. For simplicity this was not shown on the diagram. The 32 counter is at O. mode is 1, chip select will be on CS1. The position ROM outputs the coordinates of the phosphene produced when the electrode at this initial address is energized. When the image matrix receives the "store address pulse", CPM4, its internal register takes these coordinates and enables the phototransistor associated with them. Subsequently, an analog output appears at the out line and is amplified, processed and digitized by an A-D converter.
CPM4 also goes to the latches on the RAM's address lines and the output of the intensity ROM. Now, when the adder is strobed, CPM3 it adds the actual digital intensity and the ROM's intensity correction factor. This 5 bit "corrected" intensity word goes to the multiplexer's latch and into the RAM for storage. The latch is activated, CPM5 and this 5 bit word goes into the multiplexer. At the same time the 32 counter is incremented sending a new address to the ROM's. Now that the first word has gotten into the multiplexer the "start up" sequence is completed and the masterclock can be applied to all units.
The ring counter sends the intensity information out a bit at a time via the multiplexer. At the same time the image matrix is getting information for the next point, as soon as the 5 bits have been read the new intensity word shifts in. After four cycles, the shift registers of the internal packages are filled. During the ring counter's 5 state, the clock time after the 4th work, a "store" pulse is sent to activate the internal latches. When this store pulse occurs, the clock pulse being sent out is stopped for 1 clock cycle so that the 24th shift does not occur internally. Note that this 5th bit of the multiplexer supplies both D-A polarity and -the store pulses to the internal package.
The same cycle will continue until 32 words have been sent out. Now the mode controller goes to the 2 mode and cycling begins again and now information is read out of the RAM. After 32 words have been sent in this mode, the mode controller is set to 3 mode, zeroing the data to the multiplexer. This state is held for another 32 words of output. The mode controller now goes back to 1 mode and the 8 counter is incremented. All the mode cycles will now repeat for the second 32 (4x8) group of electrodes then continue to repeat until 32 groups have been completed for the third and fourth. At this point, the whole matrix of 1024 points has been accessed once; the counters are at their initial states again and the 1024 electrodes cycle repeats.
COMPONENTS
As in the internal packages the external package makes extensive use of MOS IC's (dynamic and CMOS). The ROM's are all electrically programmed MOS-avalanche types to permit programming changes, since threshold and map characteristics would vary from user to user. For the most part, the hardware consists of standard items. Because the external package is accessible for repairs it is not necessary to incorporate any redundancy into its design. This external unit could be built from about 30 IC chips and therefore would be quite compact.
COMPLETE SYSTEM
Two internal packages and one external package are interconnected as shown in Figure 10 to make up the system. The only items not within these packages are the image matrix; batteries, power oscillator and the inductive links. With the exception of the power receiver the receiver blocks shown are simply pulse shaping circuits which may or may not prove to be necessary. The power receiver is a tuned circuit resonant at 10MHz, the power oscillator frequency. A full wave Schottky diode bridge rectifies the signal and splits it into positive and negative voltages. Two small capacitors act as filter.
The image matrix would be connected to the external package through a small multi-conductor cable and could be mounted on an eyeglass frame or other fixture.
CONCLUSION
A 1024 point artificial vision system is feasible given today's integrated circuit technology. Using a time share (multiplex) arrangement, the number of inductive couplings can be minimized while still keeping the electronics compact and practical. Although the actual experience of using electrically elicited phosphenes under chronic physiological conditions is virtually limited to the work of Brindley and his collaborators, their observations coupled with related findings of others, yield a reasonable premise upon which to design a system. Furthermore, the system can be modified within limits if and when better stimulation parameters are discovered. The visual prosthesis presented here is small and should have modest power requirements, on the order of 750 mW internally and 1 watt overall.
If the growing interest in and awareness of the potential value of phosphene visual prosthesis for the blind continues at its present rate, artificial eye design such as ours may attract the necessary support to be built and tested. It would be desirable to try it first in monkeys and then in fully informed and consenting blind human volunteers. In this way, we hope to achieve our goal of providing the possibility of vision for the blind in the not too distant future.
REFERENCES
1. Foerster, 0., Beitrage zur Pathophysiologie der Sehbahn und der Sehsphare, J. Psychol. Neuro. (Lpz.) 39, 463-485, 1929.
2. Krause, F.H. and H. Schum, Die epileptischen Erkrankungen, in Ho Kuttner (ed.), Neue Deutsche Chirurgie Vol. 49 Enke, Stuttgart, 1931, pp. 482- 486.
3. Button, J.C., Jr., Electronics brings light to the blind, Radio Electronics 29 / 12, 53-55, Dec. 1958.
4.Button, J. and T. Putnam, Visual response to cortical stimulation in the blind, J. Iowa State Med. Soc., 52, 17-21, 1962.
5. Brindley, G.S., Transmission of electrical stimuli along many independent channels through a fairly small area of intact skin. J. Physiol., 177, 44-45P, 1965.
6. Brindley, G.S., The number of information channels needed for efficient reading. J. Physiol., 177, 44P, 1965.
7. Brindley, G.S. and W.S. Lewin, The sensations produced by electrical stimulation of the visual cortex. J. Physiol., 196, 487-493, 1968.
8. Brindley, G.S. , Personal communication, 1972.
9. Anon., Controversial research into new kind of sight. Med. World News, 26-34, Sept. 5, 1969.
10. Brindley, G.S. Sensations produced by electrical stimulation of the occipital poles of the cerebral hemispheres and their use in constructing visual prostheses. Ann. Roy. Coll. Surg. 47, 106-108, 1970.
11. Anon., Artificial vision: microelectronic implant for directly stimulating the brain of blind people. Wireless World, 214-217, May 1971.
12. Donaldson, P.E.K., Visual prosthesis - an implantable electrical aid for the sightless. Electronics and Power 439-440, Nov. 1971.
13. Brindley, G.S. , P.E.K. Donaldson, M.S. Falconer and D.N. Rushton The extent of the region of occipital cortex that when stimulated gives phosphenes fixed in the visual field. J. Physiol., 225/2, 58-59P, 1972.
14. Donaldson, P.E.K. Experimental visual prosthesis Proceedings, 120, 281-298, 1973., IKE
15. Brindley, G.S., The variability of the human striate cortex. J. Physiol. 225/2, 1-3P, 1972.
16. Marg, E., J.N. Fordemwalt and J. Miner Design for a phosphene visual prosthesis, Brain Res. 19, 502-510, 1970.
17. Guyton, D.L. The capacitor stimulating electrodes. Presented at the Workshop of Functional Neuromuscular Stimulation, National Institutes of Health, Bethesda, MD, April 27-28, 1972.
18. Mao, C., D. Hodges, and E. Marg Computer-controlled phosphene prosthesis testers In preparation, 1973.
19. Leonhardt, M.L. and D.A. Hodges An inductively coupled power source. In preparation, 1973.20. Reticon Corp., Mountain View California