According to the World Health Organization, about 1.3 billion people in the world have pathologies of the organ of vision, and 36 million suffer from a complete loss of visual function. The most common causes of irreversible vision loss in most developed countries are age-related degenerative eye diseases such as glaucoma, macular degeneration, cataracts, diabetic retinopathy [1–3], and trauma [2]. The consequence of blindness is a significant deterioration in the quality of life of an individual, disability, psycho-emotional disorders, and a high risk of unintentional injuries [4]. In connection with the foregoing, the problem of restoring or compensating the function of the visual analyzer in humans is very relevant. If the pathological process that led to the loss of vision cannot be corrected conservatively or surgically, the only way to restore the ability to see can be the use of a visual prosthesis — a bionic eye.
The purpose of this review is to highlight the main approaches to the development of prosthetic vision systems.
In 1775, C. Le Roy discovered the occurrence of phosphenes during electrical stimulation [5–7], and at the beginning of the 20th century. the first experiments on stimulation of the human visual cortex on the open brain were carried out [6]. In the middle of the XX century. J. Button and T. Putnam implanted two pairs of stainless steel electrodes into the visual cortex of a blind patient connected to a cortical electrical stimulator with two photocells, thanks to which she could distinguish between the presence or absence of light and determine its brightness [6, 8]. The first implantation of a multi-electrode device into the human visual cortex was performed in 1967 by G.S. Brindley and W.S. Lewin [9]. The system they created used wireless communication with the implantable part of the prosthesis, which eliminated the risk of infection [10], which anticipated many modern developments of this kind.
Basic principles of constructive implementation of visual prostheses
At its core, any visual prosthesis is a device based on a neuromachine interface, a technical system that provides direct interaction between different regions of the brain, as well as between the nervous system and external electronic devices [11, 12]. Most often, a visual prosthesis includes a device for obtaining visual information (a video camera or an array of photodiode cells), a video signal processing system, and a stimulator with an array of electrodes for signal transmission to the nervous system [13]. The above modules are structurally most often distributed between two parts: external and internal (implantable).
The outer part contains a video camera, an image pre-processing system, and a wireless interface for transmitting the processed image to the inside of the prosthesis [14]. The most common option for attaching a video camera is usually to embed it in the frame of glasses, especially in the case of anatomically intact eyeballs [14, 15]. This method is simple to implement, but has disadvantages: the impossibility of physiological positioning of the direction of the "look" of the camera, due to which the users of such a prosthesis have to perform compensatory head movements to scan the visual field, and the discrepancy between the direction of the optical axis of the eyeballs in motion and the direction of the optical camera axis. One of the solutions to these problems is the use of gaze tracking systems that track its direction and make appropriate compensatory adjustments to the image received from the camera [16]. Another way is to transfer a video camera inside the eyeball, for example, by implanting a chip with an array of photodiodes into the retina (the optical system of the eye is used for focusing) [17] or by implanting a miniature video camera with its own optical system instead of a lens [18].
The image pre-processing system allows you to optimize visual information for its presentation in the form of stimulating signals. Most often, the following operations are performed on the original image: quality correction, cropping, brightness and contrast adjustment, edge detection, image conversion from color to grayscale, segmentation, resolution reduction (to the number of pixels corresponding to the number of stimulating electrodes) [14, 19]. After pre-processing, the video signal is transmitted via a wireless interface to the interior of the prosthesis, usually including a wireless interface receiver; a controller that controls the operation of the implantable device and generates stimulating signals based on the data received from the external part; analog multichannel stimulators that generate current pulses of specified characteristics that can cause excitation in the target neural structures of the visual system and stimulating microelectrodes that are in contact with body tissues [14, 20]. The design of the electrodes largely depends on the target structure for which they are intended to stimulate, which, accordingly, determines the localization of the implantable device. On Fig. 1 shows options for implanting stimulating electrodes in visual prostheses of various types. There are retinal visual prostheses, in which the image is transmitted by means of stimulating electrodes to the remaining living cells of the retina [13], prostheses with optic nerve stimulation [21], and cortical visual prostheses, in which stimulating electrodes are implanted in the visual cortex [6, 22].
Fig. 1. Variants of arrangement of stimulating electrodes in various types of visual prostheses.
Violation of vital activity and death of photoreceptors of the retina deprive it of the ability to perceive light, which leads to gross visual impairment and loss. An example of such a disease is retinitis pigmentosa, which is characterized by a progressive course with primary rod degeneration followed by cone death, which sooner or later leads to the complete loss of photoreceptors [23]. However, even in blind patients with dystrophic diseases of the retina of this kind, its other neuronal elements and pathways of the visual analyzer remain alive, which means that their stimulation can cause the appearance of visual sensations. Electrical stimulation of the remaining more proximal retinal neurons underlies retinal visual prostheses [13]. Currently, this type of visual prosthesis is the most common. The advantages of this method are the relative simplicity and safety of surgical access to the retina for implantation of stimulating electrodes due to the extracranial location of the target structures, a simpler organization of the retina in comparison with structures further along the path of visual signal transmission [24], the possibility of stimulating retinotopically located cells, which makes it possible to provide the patient with visual sensations close to physiological [25]. The disadvantage of this method is the need the preservation of most of the structures of the visual analyzer — retinal visual prostheses are effective only in patients with diseases that affect only the very initial parts of the path of the visual analyzer, such as age-related macular degeneration and retinitis pigmentosa [25]. In the case of diseases leading to atrophy of the optic nerve (for example, glaucoma), or severe injuries with the destruction and loss of the eyeballs, visual prostheses of this type are not effective. Options for the location of implantable multielectrode matrices of retinal prostheses are shown in Fig. 2.
Fig. 2. Different arrangement of the implantable array of stimulating electrodes in retinal visual prostheses.
The relative thickness and size proportions of individual eye tunics and retinal structures here are not actual, changed for illustrative purposes.
Matrices of epiretinal prostheses are implanted on the surface of the retina in the vitreal cavity, the target neurons for stimulation are ganglion cells [24]. The advantages of this localization are the ability to use lower voltage and current to stimulate the retina due to the close location of the electrodes to its neurons, the removal of excess heat from the device through the vitreous body [13], and the absence of the need to separate the retina from the pigment epithelium and choroid [14]. The disadvantages include the usually existing need to use a special anchor nail to fix the matrix and problems with the mechanical stability of the implanted device in the long term [24]. Ganglion neurons are the output layer of the neural network of the retina, where visual information arrives already in a pre-processed form, which should be taken into account when developing algorithms for stimulation and image pre-processing [14]. An important issue is to ensure selective stimulation of only ganglion cells located under active electrodes, without involving fibers coming from ganglion cells from other parts of the retina [24]. An example of a commercially available epiretinal visual prosthesis is the Argus II system (Second Sight Medical Products, Inc., USA) [26–28].
Subretinal prostheses implanted in the space under the retina are usually implemented as an array of photosensitive cells and stimulating electrodes on a single chip, without the need for an external camera [24, 29]. Target structures for stimulation are retinal bipolar neurons [14]. The advantages of this approach are the solution to the problem of implant fixation — it is mechanically held in the subretinal space [24]; the possibility of scanning the visual space with the movement of one's own eyeballs [30]; stimulation at the level of bipolar neurons, which makes it possible to perform neural processing of the signal at the level of the retina and form the necessary patterns for the reconstruction of the visual scene by the already overlying sections of the analyzer. The disadvantages of subretinal stimulating electrode implantation are associated with the risk of thermal damage to the retina and, in the presence of a cable passing through the sclera in some models, with the risk of developing subretinal hemorrhage or complete retinal detachment in the long term [14]. An example of a commercial subretinal artificial vision system is the Alpha AMS of the German company Retina Implant AG [31].
In the suprachoroidal approach, electrodes are implanted between the sclera and choroid. The obvious advantages of such localization of the implanted part of the prosthesis are easier surgical access and less invasiveness [32]. However, the greater distance of the stimulating electrodes from retinal neurons necessitates a greater amplitude of stimulation currents and the difficulty of providing a high resolution of the resulting image [33]. Intrascleral implantation of electrodes, classified by some authors as an independent group, is inherently close to suprachoroidal, with the only difference being that the electrode matrix is located in the thickness of the sclera in a specially formed pocket [26, 34], not in direct contact with the choroid, otherwise from this arrangement, the same strengths and weaknesses of the method follow logically as in the previous one.
Implantation of stimulating electrodes into the optic nerve
When electrodes are implanted into the optic nerve, the target structure for electrical stimulation is the axons of retinal ganglion cells contained in it [21]. There are two approaches to stimulating the optic nerve: using electrodes located outside the nerve in a special cuff, and using an array of needle electrodes penetrating the nerve [24]. The first method achieves less mechanical traumatization of the nerve, however, it is inferior to the invasive approach in terms of its spatial resolution and requires higher stimulating currents [35]. In general, the advantages of the optic nerve as a place for electrode implantation are relatively easy surgical accessibility, the ability to achieve stimulation of the widest possible area of the optic fields and a relatively good visual topography correspondence of the sections of the nerve section to the sections of the retina [21, 25, 32]. An example of experimental development for the creation of prostheses using optic nerve stimulation is the C-sight (Chinese Project for Sight) project, within which a visual prosthesis with penetrating needle electrodes implanted in the optic nerve is being developed [21].
Cortical visual prostheses
In cortical visual prostheses, the target structures for stimulation are the higher parts of the visual analyzer [25], which implies the implantation of stimulating electrodes into the visual cortex of the brain [32]. This approach has a number of advantages over implanting electrodes in the retina and optic nerve. The larger surface area of the visual cortex, in comparison with the area of the retina and the cross-sectional area of the optic nerve, makes it possible in the future to implement prostheses that transmit a high-resolution signal to the brain, due to the ability to place a large number of electrodes, while the process of placing a microelectrode array on the surface of the cerebral cortex technically simpler than the microsurgical techniques required for the implantation of most epiretinal and subretinal prostheses [8]. While all retinal prostheses and prostheses based on implantation of electrodes into the optic nerve require at least partial preservation of the retina (at least ganglion cells and their processes), their use is limited to diseases that primarily affect the photoreceptor layer. In neurodegenerative diseases such as glaucoma, accompanied by the death of already ganglion cells and atrophy of the optic nerve, these types of prostheses are not applicable. At the same time, glaucoma is one of the most common causes of blindness [36], which makes it important to develop visual prostheses capable of restoring vision in patients with glaucoma. Cortical visual prostheses, transmitting visual information directly to the cerebral cortex, bypassing the initial sections of the visual analyzer, can be used in patients with severe (up to complete destruction) lesions of the retina and optic nerves [8, 22], such as glaucoma, optic nerve atrophy, injuries of the optic nerve and/or eyeballs [25].
Speaking about the advantages of cortical visual prostheses, it would be fair to note the technical difficulties of their implementation. Despite the aforementioned ease of placement of the microelectrode array on the surface of the cortex compared to its placement in the structures of the retina, the neurosurgical operation itself is a rather complicated and risky procedure [25]. A more serious problem is the development of algorithms for the formation of such stimulation patterns that would allow the formation of visual sensations that reflect the physiological visual perception of the external world, which is less relevant for the types of visual prostheses described earlier in this review. In the case of cortical visual prostheses, one of the solutions may be the use of algorithms for dynamic stimulation of the visual cortex [22]. More about the problem of perception when using visual prostheses will be discussed in the next section. An example of a cortical visual prosthesis under development for commercial use is the Orion device from Second Sight Medical Products, the developer of the Argus II system, which is currently in active clinical trials (NCT03344848) [24].
The picture perceived by patients when using visual prostheses
When developing and using visual prostheses, it should be taken into account that at present the visual scene they form, perceived by the patient, differs significantly from that provided by the operation of the natural visual analyzer in a healthy person. Single phosphenes, which appear during stimulation of small separate areas of the visual cortex, are perceived by a person as small bright points of light [37]. Stimulation of many points of the retina or visual cortex already forms a pattern of phosphenes [38]. At the dawn of the development of visual prosthetics, the opinion prevailed that, by stimulating multiple points in the visual cortex or retina in parallel, it was possible to control the pattern of phosphenes like a raster computer display, setting the values of its individual pixels and forming a continuous integral image. To date, this concept is recognized as not fully consistent with reality, especially in relation to cortical stimulation [20, 22]. If for the retina the map of phosphenes induced by the visual prosthesis has a relative visiotopicity, predetermined by the correspondence of the stimulation pattern to the images projected onto its surface, then in the optic nerve there is a redistribution of the relative arrangement of fibers, due to which, when the visual cortex is stimulated, phosphenes have a chaotic arrangement [20]. This poses a separate task for developers to map phosphenes and appropriately modify visual cortex stimulation algorithms to enable the patient to perceive an image that is more consistent with the whole picture.
In addition to the morphophysiological features of various parts of the visual analyzer, the limitations are imposed by the technical implementation of the prostheses themselves. The main problem is the significantly smaller number of available stimulation channels than is necessary for the formation of comfortable visual acuity: the patient perceives a much smaller amount of visual information; the scene has a low resolution, does not convey information about color and depth; no binocular vision. The reason for the above is clearly demonstrated by the following example: the Argus II retinal prosthesis has only 60 stimulating electrodes, Alpha IMS — 1.5 thousand, however, the visual analyzer contains 130 million photoreceptor and 1.3 million ganglion cells [19], which indicates the need for significantly more channels of stimulation.
Conclusion
Visual prostheses have a half-century history of active experimental development of technology. To date, several approaches have clearly taken shape, differing primarily in the target structure, the stimulation of which is necessary for the formation of visual sensations — these are retinal prostheses, optic nerve stimulation, cortical visual prostheses, etc. Stimulation of the surviving neural structures of the retina is more an easy task and currently the most developed: there are commercially available retinal prostheses, but they have a narrower scope, which is limited to diseases associated with impaired activity of predominantly photoreceptors. The development of cortical visual prostheses is more difficult than retinal ones, but in the future they may give more promising results, since in the future they allow the use of a larger number of stimulation channels to obtain more detailed visual perception, and they are also more versatile, since they do not need to preserve any elements of the visual analyzer, except for the primary visual cortex. At the moment, one of the key global tasks facing the developers of visual prostheses is to increase the realism and physiology of the visual sensations formed in patients.
Author contributions:
Research concept and design: S.K., S.S., V.M.
Text writing: S.K.
Editing: S.K., V.M.
The authors declare no conflicts of interest.