Fig. 1. External circuits. For explanation, see text.
Neurosurgeons have long known that if the visual cortex of a conscious patient is stimulated electrically, a spot of light termed a phosphene is reported localized in his visual fields (8,9,11). A crude prosthetic device employing phosphenes was pioneered with the implantation of electrodes in the visual cortex of 3 blind patients by Button and Putnam (6). Two of them were able to find the position of a flashlight by scanning manually with a photocell which provided a signal directly to the visual cortex.
Leads going through the skin into the brain, as in the Button and Putnam device always risk an infection reaching the brain through the open wound. Electromagnetic or other wave transmission through the intact skin, to which radiation the tissue is transparent, would avoid this dangerous possibility. Brindley (2,3) 3 and Brindley and Lewin (5), after some tests in baboons, have implanted under the scalp of a blind person an array of 80 radio receivers with leads to an array of 80 electrodes on the visual cortex. They showed that such a device works in principle because:
(1) Punctate white phosphenes are seen in apparently fixed positions in the central visual field as a result of electrical stimulation of the visual cortical surface through a platinum electrode (about 1 mm^2). Similar small, but non-punctate phosphenes appear in the periphery.
(2) With moderate stimulus strength, the phosphene ceases when the stimulus is turned off.
(3) The device appears well tolerated by the tissue and is still in place and functions more than 2 years after the original implantation (4).
(4) Figures can be formed and recognized such as a number 7 or a ? but the device has not proved a useful prosthesis because there are not enough functional electrodes well enough placed to form phosphene images of all alphanumeric characters and other objects of visual interest (4).
To be thoroughly useful, the prosthesis should be operable 16 h a day. The limitation, if any, of the amount and duration of stimulation that may be applied before the potentially deleterious effects of electrolysis causes damage to the tissues is unknown. This problem must be attacked experimentally with electrodes functioning in situ because it is largely dependent upon blood, cerebrospinal and tissue fluid circulation,, which is virtually unmeasurable, in the vicinity of the electrode surfaces, where heat is generated and electrolytic products are formed.
While these problems are being formulated and tested, there is time to consider the design of the prosthetic device itself. The successful stimulus parameters of Brindley and Lewin (5) can be assumed until modified by further research in the future. Their results indicated that many more electrodes would be necessary but the use of discrete components in their design such as transistors, resistors and capacitors because of their volume limits the number of channels.
The development of metal oxide silicon (MOS) integrated circuits - now widely used in electronic desk calculators and computers - significantly reduces the volume of the electronic circuits. Digital techniques can be employed, further increasing efficiency, and redundancy in design can increase reliability with no significant increase in volume. New problems of packaging the microcircuits in a body tissue environment are introduced but these problems appear solvable. We offer this design as a preliminary one, subject to modification as more data become available on the electrical generation of visual phosphenes and their use as a practical visual prosthesis. An earlier version of this design has been discussed (10) and a design with some similarities to that one has been proposed independently by Vaughan and Schimmel (12).
The design
General: We will take the following reasonable values based on the work of Brindley and Lewin (5). The surface area of the platinum electrodes is about 1 mm^2 and has a resistance of 3,000 Mohm. Maximum stimulus voltage is 20-30 V; pulse duration 100 (up to 1,000) usec; and repetition rate 100/sec.
Aside from possible special image processing devices to be added later, the prosthesis consists of two parts, external and internal. The external circuits including rechargeable batteries will lie on top of the head within (the space of) a man's hat. Internal circuits will be implanted in a hole in the cranium over one hemisphere, a short cable leading to the array of electrodes fitted to the surfaces of the visual cortex of the ipsilateral hemisphere. The skin of the incision will heal and the power and signals will be electromagnetically coupled through the scalp.
In our design, we decided to use 512 electrodes. This number was arrived at as the result of several considerations: 512 is 2^9, digitally this means that all 512 electrodes can be addressed using a 9 bit length word in the decoding; the package problems of handling 500 or more leads seem formidable enough without going any larger (the next logical number is 2^10 or 1024, and 1000 or more leads from a single package seem to be at or beyond the state of the art today). Also, since one MOS output transistor and associated bonding pad are necessary for each electrode, the simple size of the MOS devices and wiring complexity intuitively dictate that the smaller number of electrodes is a more reasonable goal. The even more reasonable number of 256 electrodes was discarded as probably not sufficient to produce a usable prosthesis. It is felt for the moment that it will prove simpler to use more packages for each person (at least one for each hemisphere) than to increase the number of electrodes in a package. As the state of the art progresses, it may well become practical to increase the number of electrodes in a package.
External circuits: Fig. 1 shows the proposed electrical circuits that are required external to the skin implant. Objects to be visualized are focused by a lens (such as a commercial 8 mm zoom lens) on a light sensitive matrix 16 by 32 made from MOS transistors on a single silicon chip. (A number of electronics manufacturers both in this country and abroad are working on such solid state image converters (1).) The matrix is scanned by a 9 bit counter whose output is the address of the matrix location being sensed.
Fig. 1. External circuits. For explanation,
see text.
It is well known that there is a lack of strict isomorphism between loci in the visual field and those in the visual cortex. The problem is compounded because electrode arrays are usually flat on the surface and cannot follow in the grooves which are formed by the internal surfaces of the sulci. For these reasons, isomorphically arranged matrix addresses may not correspond to the required visual cortex loci. Therefore the addresses pass through a position translation Read Only Memory (ROM) which can be programmed to convert any matrix address to any visual cortex address required to put the visual field in spatial registration with the visual cortex array. In use, it will be necessary first to organize the spatial array using a patchboard arrangement instead of a ROM, with the patient indicating the location of a phosphene on his visual field for a given electrode address. Using the data thus generated, a matrix of positional translations will be organized. Then, once the translational matrices are determined for a specific patient, an MOS ROM will be fabricated for this person. Each ROM will be unique for a particular patient.
A somewhat similar problem exists in regard to threshold and suprathreshold stimulus intensities. Each locus on the cortex has its own individual threshold. The electrical stimulus, the analog signal for which comes through the buffer amplifier and is converted to a digital signal of 3 bits, must be adjusted so that it is neither below threshold nor too far above it. The former would give no phosphene whereas the latter would give perseveration of the response after cessation of the stimulus, and in some instances multiple phosphenes or possibly pain. Between these two limits the intensity can be adjusted to give a limited grey scale. For these reasons the visual cortex address passes through an intensity weighting ROM which can be programmed to increase or decrease the stimulus intensity of any electrode to the desired threshold level. The intensity ROM output is added to or subtracted from the actual intensity which is obtained from the 3 bit analog to digital converter. It is assumed that grading the stimulus intensity giving a grey scale, though limited in range, will be of value.
The intensity adder feeds into a holding register which sends the final intensity code into the intensity coil pulse generator which is triggered by the advance counter for each change of matrix position. The intensity information is transmitted as a 4 bit signal by the intensity coil across the skin to the internal intensity coil which will be considered later.
Returning to the position translation ROM, it also feeds into a parallel to series converter whose output governs the two pulse generators which transmit the position information over the zeros and ones coils. Each time a position is counted in the first 9 bits in the ring counter, the 10th bit generates the advance signal which introduces the next photomatrix position as well as triggers the intensity coil pulse generator as mentioned above and the store coil pulse generator, transmitting a signal to the internal store register.
Each position bit is released from the parallel to series converter in 2 usec intervals, requiring 18 usec for the 9 bit position or 20 usec including the advance counter signal. This allows approximately 500 electrodes to be energized in 10 msec or each electrode energized at 100 Hz.
Triple redundancy can be incorporated so that the electrode is energized only when at least two of the three parallel circuits are activated.
Power is transmitted from a power oscillator at approximately 10 MHz.
Internal circuits: Pick up coils under the scalp obtain power and signals for intensity, store and position in terms of separate binary zeros and ones (Fig. 2). Sealed in silastic rubber, the coils must be as close to the skin as practicable in order to achieve good coupling.
The power is rectified and filtered, providing +- 15 V for the logic circuits. Zeros and ones are separately channeled to a shift register, which information, when completed as signaled by a pulse via the store coil, is deposited in the storage register. The 9 bit address is next decoded and its probe is switched on so that the intensity coil provides the stimulating current. As in the external circuits, triple redundancy can be introduced to improve reliability and decrease the possibility of spurious signals.
The electrical stimulus has a biphasic waveform. A rectangular negative pulse is followed by a slower positive pulse of smaller amplitude so that the net charge transfer in each cycle approaches zero. This will, of course, minimize electrolytic products some of which are toxic.
Fig. 2. Internal circuits. For explanation,
see text.
The internal package
The internal device package must provide minimal tissue reaction and five major physical functions: (1) a long-lasting, reliable, hermetic environment for the active MOS integrated circuits*; (2) a substrate onto which the active devices may be bonded; (3) interconnect paths between devices; (4) hermetic interconnections between the sealed devices and the electrode array; and (5) attachment pads suitable for attaching the leads from the electrode array.
* Sodium ion which is abundant in tissue fluids destroys MOS integrated circuits by setting up parasitic surface conduction channels and degradation of other device parameters.
The number of leads and interconnects (in excess of 500) in the necessarily small package requires a combination of a number of high density circuit substrate techniques. Each of these methods has been successfully utilized; however, the combinations of these techniques do not exist in any present package form or with any number of leads approaching 500.
The package materials selected for cranial implantation must not interfere with the semiconductor devices or cause any significant tissue reaction. Indeed, they must not be degraded themselves by their environment or their guest devices. The package must reliably seal the devices from the destructive tissue fluids. Several materials have been used for package insulating substrates and conductive circuit patterns. Among the insulating substrates are a host of plastics, high purity alumina and other ceramics. There are other insulator materials in use; however, they have not been proven in high density packaging applications. The plastics are not noted for their long term reliability when associated with semiconductor devices even in a less severe environment than the body. They are not as hermetic as are the ceramic substrates. Most military and space requirements and specifications preclude their use. Although beryllia an excellent thermal conductor, exceeding alumina in that regard, lead bonding technology has not been developed to the high strength level and reliability associated with alumina. Multilayer technology is non-existent with beryllia systems. Beryllia may also present some toxicity problems. Alumina ceramic package technology is well developed for high density, high reliability applications and is the sole choice for the development of this package.
The metal conductor systems used with alumina in package applications are many, including molybdenum, tungsten, chromium, nichrome, gold, platinum, nickel, copper, and combinations of these materials. Multilayer technology, if used, restricts the choice to platinum, tungsten, molybdenum, and nickel with added gold plating. In systems where ionic solutions are present, a single metal system is preferred to eliminate the corrosive or voltage generating effects of metallic couples (Seebeck or Peltier effects). As a first choice, it appears that an all platinum system would be most desirable and will be used, provided that adequate hermeticity can be reliably achieved. A second choice could be a molybdenum interconnect system with gold plated pads for bonding devices and platinum harness wires. The ultimate selection of materials would be based on their compatibility with the package concept as well as their compatibility with body tissues.
The number of interconnects almost precludes the use of peripheral leads as are commonly used in monolayer and multilayer packages today. We are proposing a metallized feed-through technique utilizing ceramic coated platinum wires. In principle this can provide at least 2,500 conductive vies per square inch of package and perhaps as many as 10,000 vies per square inch. The via density can be developed with solution metallizing using molybdenum technology or with cofired platinum technology. Each is compatible with multilayer circuitry and will provide adequate room and bonding pads for devices on the inside and platinum wire harness on the exterior. The selection of a metallurgical system to be used with the high density feed-through technique will be made after a final design and development study.
An artist's concept of the package (without the coils which would be external to the package) is shown in Fig. 3. A flange may be added on the outer surface with screw holes so that the box may be securely fastened to the surface of the cranium. The surface materials, alumina ceramic, platinum, Dacron and medical silastic rubber are all known to be well tolerated by body tissues.
Fig. 3. Artist's concept of internal
package in situ. Ceramic box approximately 1 in. * 2 in.
* 3/8 in. lies in a hole in the skull. Five hundred electrode
lead wires and the platinum electrodes are in a Dacron net
covered with medical silastic rubber. The latter is seen as a pad
under the ceramic package and covering the medial surface of the
visual cortex (striate cortex or area 17 of Brodmann). The
induction coils which are not shown would be in a flat silastic
mat to be placed near the skin towards the top of the head.
The electrode array
The ceramic coated platinum wire can be woven into a Dacron mesh matrix on which the electrode array is formed. The whole array can be formed in the shape of the cortical surface and impregnated with silastic rubber (Fig. 3), leaving only the electrode surfaces exposed. The electrodes would be formed of 1 mm^2 pieces of platinum, similar to those of Brindley and Lewin (5). Spacing initially will be about 1 mm between edges which may be a somewhat smaller separation than necessary for maximum phosphene resolution. Until more information is available, it is better to err in the direction of having the electrodes too closely spaced rather than the reverse. The large surface area is required to minimize the impedance. Various methods could be used to increase the surface area of the electrode by platinum black or irregular surfacing but these would erode in time, changing the electrical characteristics (7).
The visual cortical surface area (excluding the depths of sulci) of one hemisphere in man is measured at 9.7 cm^2 for area 17, 11.5 cm^2 for area 18, and 17.7 cm^2 for area 19. About 25 electrodes can be accommodated per cm^2 Initially it would be reasonable to confine the electrodes partly to area 17 of Brodmann which will accommodate about 250 electrodes, and partly to the border regions at areas 17 and 18 which include the vertical meridian of the visual field. Thus 500 electrodes should be more than adequate for the initial model. Ultimately the total surface area can accommodate up to a thousand electrodes of the size and spacing mentioned above.
The return circuit is through a strip of platinum foil implanted in the scalp. While it would be more convenient to have this foil on the back of the electrode array shell, it is probably better to have the electrolytic products formed at this surface in the highly vascular scalp where they could be disposed of more readily.
School of Optometry, University of California, Berkeley, Calif. 94720, and (J.N.F. and J.M.) American Microsystems Inc.. Santa Clara. Calif. (U.S.A.)
ELWIN MARG, JAMES N. FORDEMWALT, JAY MINER
l ANON., Array of dual-gate MOS transistors successfully generates tv picture, Electronics, Sept. 29 (1969) 199-200.
2 BRINDLEY, G. S., The number of information channels needed for efficient reading, J. Physiol. (Lond.), 177 (1965) 44-45P.
3 BRINDLEY, G. S., Transmission of electrical stimuli along many independent channels through a fairly small area of intact skin, J. Physiol. (Lond.), 177 (1965) 44 46P.
4 BRINDLEY, G. S., Report to the conference on visual prosthesis. In T. STERLING, E. BERING AND S. POLLACK (Eds.), Proc. Second Conference on Visual Prosthesis, (University of Chicago, June, 1968), Academic Press, New York, in press.
5 BRINDLEY, G. S., AND LEWIN, W. S., The sensations produced by electrical stimulation of the visual cortex, J. Physiol. (Lond.), 196 (1968) 479-493.
6 BUTTON, J., AND PUTNAM, T., Visual responses to cortical stimulation in the blind, J. Iowa State med. Soc., 52 (1962) 17-21.
7 DOTY, R. W., Outline of techniques for electrical stimulation of cortical and subcortical structures. In T. STERLING, E. BERING AND S. POLLACK (Eds.), Proc. Second Conference on Visual Prosthesis, (University of Chicago, June, 1968), Academic Press, New York, in press.
8 FOERSTER, O., Beitrage zur Pathophysiologie der Sehbahn und der Sehsphare, J. Psychol. Neurol. (Lpz.), 39 (1929) 463-485.
9 KRAUSE, F., UND SCHUM, H., Die epileptischen Erkrankungen. In H. KUTTNER (Ed.), Neue Deutsche Chirurgie, Vol. 49a, Enke, Stuttgart, 1931, pp. 482-486.
10 MARG, E., Contributions to the conference on visual prosthesis. In T. STERLING, E. BERING AND S. POLLACK (Eds.), Proc. Second Conference on Visual Prosthesis, (University of Chicago, June, 1968), Academic Press, New York, in press.
11 PENFIELD, W., AND RASMUSSEN, T., The Cerebral Cortex of Man, Macmillan, New York, 1952, pp. 135-147.
12 SCHIMMEL, H., AND VAUGHAN, H. G., JR., Feasibility of visual electrocortical prosthesis. In T. STERLING, E. BERING AND S. POLLACK (Eds.), Proc. Second Conference on Visual Prosthesis, (University of Chicago, June, 1968), Academic Press, New York, in press.