Hearing Restoration through Optical Wireless Cochlear Implants

Abstract

In this chapter, we present two novel optical wireless-based cochlear implant architectures: (i) optical wireless cochlear implant (OWCI) and (ii) all-optical cochlear implant (AOCI). Both the architectures aim to decisively improve the reliability and energy efficiency of hearing restoration devices. To provide design and development guidelines, we document their main components, discuss the particularities of the transdermal optical channel, and provide the analytical framework for their accurate modeling. Building upon this framework, we extract closed-form formulas that quantify the communication, the stimulation, and the overall performance. An overall comparison of OWCI and AOCI, as well as conventional cochlear implants, accompanied by future research directions summarizes this chapter. Our findings reveal that both the OWCI and the AOCI outperform conventional cochlear implant approaches; thus, they are identified as promising architectures for the next generation of cochlear implants.

1. Introduction

The healthy ear functions much like a receiver (Rx) of acoustic signals, which can be described as time-varying pressure waves in a specific frequency range (20–20,000 Hz). These signals propagate toward the cochlear, which analyzes them based on their spectral content. Specifically, each pressure wave traveling inside the cochlea not only actuates inner and outer hair cells at different locations along its length based on the frequency components of the wave, but also determines the intensity of the perceived sound according to the amplitude of the wave [1]. The various spiking characteristics of the spiral ganglion neurons, such as spike rate, number, and location, encode the amplitude and frequency of the sound.

The most common sensory defect is hearing loss, which plagues more than 466 million people around the world and is mostly caused by cochlear abnormalities [2]. When unaddressed, hearing loss can negatively impact the quality of life in various ways, such as social isolation, limited education, and unemployment, which are estimated to cost 980 billion dollars annually. To counterbalance this, substantial research effort has been directed toward neuron regenerative techniques, such as pharmacological, gene, as well as cell therapies [3, 4]. Unfortunately, none of the aforementioned approaches is considered to be close to clinical use. Therefore, the most successful hearing restoration approaches to this day are based on cochlear implants (CIs). Of note, CIs can be used in almost all forms of hearing loss.

Conventional CIs are comprised of two parts: one external and one implanted. The former houses a sound receiver and the processor, while the latter contains the stimulation unit. Specifically, the captured sound signal is decomposed to its major frequency components that are assigned to the corresponding channels of the stimulation unit. Each channel delivers the electrical stimulation signal to the spiral ganglion neurons that match the frequency content of the decomposed electrically encoded sound signal. However, due to the relatively high electrical conductivity of the cochlea, the applied electrical stimulation spreads to nearby spiral ganglion neurons, thus stimulating wider spectral windows than the appropriate one. In conjunction with their low-dynamic range [5], conventional CIs offer limited spectral and intensity sound encoding, which is proven to be detrimental for their hearing restoration capabilities [6].

In this chapter, we introduce the major advances that paved the way for the revolution of CIs and the realization of hearing restoration. Initially, we investigate the current state of the art of hearing restoration through CIs. Next, an in-depth analysis of most promising techniques of light-based hearing restoration is presented. Finally, we offer design guidelines as well as future directions for the next generation of CIs.

2. Background

To aid the reader in understanding the requirements of hearing restoration, we provide some background that covers the CIs’ evolution since their conceptualization as well as the current research progresses toward the next generation of CIs (Figure 1).

Figure 1.

The evolution of CIs from 1982, when the first CI manufacturing company was founded, until the current state-of-the-art research that validated the feasibility of optogenetics-enabled optical CIs.

2.1 Evolution of CIs

The concept of hearing restoration through the electrical stimulation of the auditory nerve was conceived by André Djourno and Charles Eyriés in 1957. In their attempt to restore the functionality of the facial nerve through electrical signal applied via a wire, the deaf patient experienced auditory sensations [7]. Based on these findings, multiple attempts were made around the world to develop the first CI with William House performing the first implantation in 1972 [8, 9]. Moreover, the first cochlear implant manufacturing company was founded in 1982 under the name MedEl Corporation, closely followed by Cochlear Limited in 1984, and Advanced Bionics in 1996.

Since their creation, CI companies have iteratively updated their architecture designs, hardware, and optimizing stimulation techniques. The first generation of CIs was released in the early 1980s and included Nucleus 22 and Comfort CI, combined analog signal processing strategies with a multichannel stimulation unit that housed 22 and 4 channels, respectively. These designs were followed by the initial model of Advanced Bionics called Clarion in 1996 that was encased in a ceramic case, contained eight channels, and used rechargeable batteries. The second generation included Clarion II, Nucleus 24 Contour, and Combi 40+. These were introduced in the market with 24 electrodes and new sound processors with novel features such as precurved electrode arrays, backward compatibility, frequency modulation capabilities, dual electrodes, and behind-the-ear external components. However, in the early 2000s, completely redesigned highly customizable CI models, namely, Freedom, Pulsar, and HiRes90k, were developed. Their modularity and customization options were the distinguishing factors for these new models that were available in straight or precurved, standard, medium, condensed, and split electrode array architectures, based on the individual particularities of the cochlea of each patient. Moreover, these electrodes were encased in flexible plastic and housed a plastic tip that enabled nontraumatic implantation. In the era after 2010, the latest iterations of CIs have been focused toward higher fidelity sound that enhances the perception of music through state-of-the-art sound processing, wireless control, and software-enabled programability, as well as waterproof designs.

2.2 What the next-generation CI should be?

The utilization of light-based communications and stimulation has been proposed as a promising alternative for electrical hearing restoration techniques. The superior communication performance of optical wireless communications in trandermal applications revealed the benefits that can be achieved by utilizing light for the communication between the external and implanted components of CIs [10, 11]. Moreover, optogenetics was initially reported by Izzo et al. [12] and has been proven achieve more efficient coding of the spectral information of sound due to its higher temporal confinement compared to the electronic stimulation techniques [13, 14, 15]. Although optical stimulation has great potential, it exhibits increased energy requirements for achieving the actuation of spiral ganglion neurons, and thus, future research is necessary for developing more energy-efficient techniques [16]. Finally, the combination of optogenetics and optical wireless communications offers great promise for the realization of an all-optical CI architecture capable of achieving unprecedented performance [17].

3. Current research progress

Two main research directions remain to be investigated. First, transdermal communication plays an important role in propagating the sound information captured by the microphone of the external component toward the one implanted. Conventional CIs are based on magnetic coupling, a near-field technique that uses low radio frequencies (RFs) in the range of 5–50 MHz for communication [18, 19]. The required power of conventional CIs lies around some decades milliwatts. Although this technology has been successfully applied in the majority of CIs, it suffers from low data rates, which constrain the performance of artificial hearing aids in their attempt to simulate high-quality normal hearing [20, 21, 22]. In addition, the aforementioned spectral window is used by numerous applications, which generate a great amount of interference that diminishes the quality of communication [23, 24, 25]. On the other hand, the optical activation of the auditory nerve via optogenetics has been experimentally verified, but the propagation of the spiral ganglion neuron potential through the auditory pathway toward the brain and its successful perception have yet to be demonstrated [26]. Moreover, the superiority of optical over electrical cell stimulation must be validated in order to justify the research effort toward the all-optical cochlear implant (AOCI) [17]. Recently, multiple experiments have progressed these goals by implanting novel tiny optical fibers in animals models of human sensorineural hearing loss [27, 28, 29].

3.1 Communications

To overcome the aforementioned CI restrictions, researchers have investigated the viability of transdermal wireless networks that operate in nonstandard frequencies. Owing to increased bandwidth, surprisingly high tolerance to external interference, and partial skin transparency at near-infrared wavelengths, optical wireless communications have been applied to transdermal channels instead of the traditional RF-based techniques [30, 31]. In the past decade, numerous contributions have experimentally verified the practicality of transdermal optical links [32, 33, 34, 35, 36]. Abita established a transdermal optical link from the inside toward the outside component of a medial system achieving high-data-rate communications [32], while Ackermann et al. investigated the design principles and tradeoffs that are entangled to optical-based CIs [33, 37]. Moreover, Liu introduced a high-data-rate transdermal optical link for implantable biomedical systems with high energy efficiency under the assumption of deterministic misalignment [24]. Similarly, the interactions between data rate, transmission power, receiver characteristics, and tissue thickness as well as their impact on the system’s performance were evaluated for transdermal optical links applied in neural signal extraction scenarios [38]. In addition, the same authors validated the proposed system by conducting in vivo experiments that achieved $2\times {10}^{-7}$ bit error rate (BER) and 100-Mbps data rate under stochastic misalignment, but with relatively high power consumption in the order of 2 mW [36]. On the contrary, a novel retroflective architecture was presented for transdermal optical links [34], while Liu proposed a bidirectional transdermal optical link [35].

Building upon the aforementioned contributions, the development of optical-based CIs needs to leverage breakthrough technologies while taking into consideration the particularities of the transdermal and in-body optical channels, the space and energy design limitations, as well as the directionality of the optical links. Moreover, a novel information-theoretic framework is required for the design of energy-efficient physical and medium access schemes, as well as the development of simultaneous light information and power transfer policies and resource allocation strategies. Motivated by the above, recent research effort has been devoted toward delivering safety and high quality of experience in CIs and identifying the critical technology gaps and the appropriate enablers.

3.2 Neural stimulation

After communicating the information from the external environment toward the implanted component of the CI, the techniques of neural stimulation must be applied in order to excite the cochlear spiral ganglion neurons, which, in turn, will generate the desire action potential that will propagate along the acoustic nerve toward the brain. Over the years, various methods of neural stimulation have been developed. These can be categorized based on the nature of the applied stimulus as acoustic, thermal, magnetic, chemical, optical, and electrical, with the last ones being the most recognized [39]. Specifically, electrical neural stimulation is the most common technique and has been used in a wide gamut of biomedical applications [40, 41, 42]. Electrical neural stimulation applies an electrical stimulus (voltage, current [40], or charge [42]) on the target nerves that manipulates their membrane potential so that it exceeds a certain threshold and, therefore, generates or inhibits action potentials. Specifically, deep brain and cardiac muscle stimulation techniques that use voltage control mechanisms have been investigated with regard to power consumption [41, 43], while current-controlled electrical neural stimulation in CIs is characterized by power waste in the tissue that leads to limited longevity and tissue damage [40]. Voltage-controlled electrical neural stimulation is proven to be more power efficient and less complex, but with very limited stimulus tuning options that result in faster degradation of the electrode contacts. The opposite is valid for current-controlled electrical neural stimulation that can apply fine-tuned charge to the electrodes but exhibits lower power efficiency. Finally, charge control mechanisms for electrical neural stimulation have been applied on the peripheral neural system [42] and offer a middle ground between stimulus control and power consumption. Despite the control mechanism, the determining factors of electrical neural stimulation techniques include human safety, energy efficiency, stimulation waveform, and spatial resolution. The latter significantly affects the stimulation accuracy and is correlated with the distance from the targeted neurons as well as the size of the electrode, which is limited by maximum permissible charge per tissue surface and the electrode’s manufacturing process. In addition, the unique characteristics of different types of neurons greatly affect their response to stimulations with variable waveform properties, such as amplitude, width, and frequency. To this end, a great amount of research effort has been devoted toward optimizing the waveform for the stimulus [44, 45, 46, 47]. Finally, throughout the optimization procedure of electrical neural stimulation techniques, safety for humans must be ensured.

The solution to the several limitations of electrical neural stimulation was introduced almost two decades ago in the form of optical neural stimulation that uses light for the actuation and control of neurons. Specifically, light-gated ion channels found in proteins, termed opsins, have been proven to mediate light-driven action potentials in mammalian neurons by manipulating the polarization of their membrane and, therefore, suppressing or exciting them. Optical neural stimulation is highly dependent on the type of the utilized opsin, which incentivized research toward experimentally verifying its performance in terms of precision, accuracy, frequency, and scalability [48, 49, 50, 51]. Optical neural stimulation was successfully applied in the motor control system of rodents [48], while the causal relationship of frequency-based optical neural stimulation and behavior state transitions was verified [49]. The increased specificity of exciting neurons was illustrated through efficiently mapping the spatial distribution of synaptic inputs [50]. Moreover, a high-precision optical neural stimulation technique for inhibiting neurons with temporal fidelity was developed [51]. The performance of this technique was evaluated based on novel key performance indicators such as light sensitivity. The aforementioned works illustrate that the development of opsins offering stable performance over multiple stimulations is accompanied by long desensitization periods and short channel-off durations. To this end, research was intensified toward developing opsins with different kinetic features and wavelength sensitivity for monitoring and controlling biological processes in subcellular and cellular levels [52, 53]. A major breakthrough was achieved with the application of channelrhodopsin 2 (ChR2) in mammalian neurons that enabled accurate stimulation with light pulses [54]. Since its development, ChR2 has been heavily investigated, and multiple variants have been introduced with applications in cardiology [55, 56, 57, 58] and neuroscience [59, 60]. The performance of these variants greatly outperforms electrical neural stimulation in terms of stimulation pulse intensity and frequency (up to 200 Hz), as well as the ability to trigger large current action potentials with higher fidelity [61, 62].

4. Light-based hearing restoration

Based on the literature review presented in the previous section, the main bottlenecks of CIs are low accuracy and low precision of nerve stimulation methods, bandwidth scarcity and constraint capacity of RF communication techniques, and high energy consumption of both. To this end, we present two architectures capable of mitigating the effect of these limitations and even eliminating them [11, 17].

4.1 Optical wireless cochlear implant

The utilization of optical wireless communications in order to develop CI transdermal optical links has been recently investigated [11], where the authors proposed a novel system architecture, termed optical wireless cochlear implant (OWCI), that improves the power and spectral efficiency as well as the reliability of the transdermal optical link. Moreover, in the same contribution [11], the capabilities and feasibility of the OWCI are evaluated and design guidelines are provided. The main comparison points between OWCIs and conventional CIs are illustrated in Table 1. In addition, the presented advances in the communications of CIs are in line with optical neural stimulation advances on the acoustic nerve [21, 63, 64, 65, 66, 67].

OWCIs	Conventional CIs
Increased data rate	Low data rate
Abundant bandwidth	Limited bandwidth
High power efficiency	Low power efficiency
Safer for the human body	Questionable safety
Mature technology with promise of	Mature technology
higher performance in the same scale	with compact designs
Low solar and ambient light interference	Very high electronic interference
Stringent alignment requirements	Susceptible misalignment
Multiple design guidelines	Mature standardization
(IrDA, EU COST 1101, IEC LSS, IEEE Std 1073.3.2-2000, etc.)	(IEEE Std 1073.3.2-2000, IEEE 802.15.4, etc.)

Table 1.

OWCIs versus conventional CIs (bold fonts demonstrate the advantages).

The unique technical contributions of the OWCI entail the establishment of a novel system model for transdermal optical links that incorporates the various design variables such as the stochastic misalignment between the receiver (RX) and the transmitter (TX), the scale of the optical components, the skin thickness, and the transmission power. The external component of the OWCI is comprised of a microphone, the TX, and a digital signal processing (DSP) unit, while the implanted one contains the RX as well as a stimulation and a DSP unit. The external DSP unit is responsible for digitizing and compressing the sound signal from the microphone into coded signals, which are then forwarded from the TX to the RX over the transdermal optical link. In the implanted component, the DSP and stimulation units transform the received signal into a series of electrical pulses that will stimulate the auditory nerve (Figure 2). Based on this system model, the performance of the OWCI was evaluated with regard to the SNR, channel capacity, outage probability, and spectral efficiency. The results not only validated the feasibility of the proposed architecture and provided meaningful insights that can be used as design guidelines, but also revealed the superior effectiveness and reliability of the OWCI compared to the conventional CI.

Figure 2.

Diagrammatic illustration of the architecture of the OWCI. The OWCI captures the sound information via the microphone located outside of the human body. Afterward, it utilizes optical wireless communications to transfer it toward the receiver fixed on the cranial bone. Finally, the implanted unit stimulates the acoustic nerve by delivering the appropriate signals via the stimulation electrode.

In the aforementioned architecture, the transmitted signal, $x$, is conveyed over the wireless channel, $h$, with additive noise $n$. Thus, the received signal can be written as [68, 69, 70].

$$y_1=\mathit{Rhx}+n$$E1

with $\eta$ denoting the quantum efficiency of the photodiode, $R$ the RX’s responsivity, $v$ the frequency of the photons, $q$ the electron charge, and $p$ the Planck’s constant. It is highlighted that the channel coefficient can be expressed as $h=h_l{h}_p$, where $h_l$ represents the deterministic channel coefficient caused by propagation loss, while $h_p$ denotes the collected power fraction due to the geometric spread from the origin of the detector and is caused from the TX-RX misalignment (Figure 3).

Figure 3.

The effect of misalignment with regard to skin thickness. As the RX (photodiode) moves away from the TX under the same severity of misalignment, its distance from the perfect alignment conditions becomes enclosed in the TX’s beam width. This phenomenon creates an equilibrium between the optimal TX-RX distance and the TX beam width under fixed misalignment conditions.

The CI channel’s deterministic term can be expressed as in ([71], Eq. (10.1))

$$h_l=exp\left(-\left({\mu }_{\alpha }\left(\lambda \right)+{\mu }_s\left(\lambda \right)\right)\delta \right)$$E2

where $\lambda$ is the transmission wavelength, $\delta$ is the skin thickness, ${\mu }_{\alpha }\left(\lambda \right)$ is the skin attenuation coefficient, and ${\mu }_s\left(\lambda \right)$ is the skin scattering coefficient, which can be acquired from a plethora of experimental results [72, 73, 74, 75, 76]. In this analysis, the term skin refers to the biological structure that consists from the stratum corneum, the epidermis, and the dermis [71], while both the RX and the TX touch with the inner (adipose) and outer (epidermal) side of the skin [24], and thus, the TX-RX distance is regarded equivalent to skin thickness.

The misalignment between the TX and the RX can be modeled as the stochastic channel coefficient, which expresses the collected power due to geometric spread with radial displacement $r$ from the origin of the detector and can be written as

$$h_p\approx Aexp\left(-\frac{2r^2}{w_e^2}\right),$$E3

which is based on the assumption that at distance $d$ from the TX, the circular aperture of the transmitted beam has a radius of $R$ and the spatial intensity on the plane of the RX is $w_d$. In addition, $w_e$ represents the equivalent beam waist radius and $A$ expresses the collected power under perfect alignment. This approximation has been utilized in various previous works for modeling stochastic pointing errors [77, 78].

Based on this model, if we assume independently and identically Gaussian distributed horizontal and vertical displacement, it has been proven that $r$ follows a Rayleigh distribution [79]. As a result, the probability density function (PDF) of the stochastic term of the channel coefficient can be written as

$$f_{h_p}\left(x\right)=\frac{\gamma }{A^{\gamma }}{x}^{\gamma -1},0\le x\le A,$$E4

where

$$A={\left(erf\left(\beta \right)\right)}^2,\beta =\frac{\sqrt{\pi }R}{\sqrt{2}{w}_d},\gamma =\frac{w_e^2}{4{\sigma }^2},w_e^2=w_d^2\frac{\sqrt{\pi }erf\left(\beta \right)}{2\beta exp\left(-{\beta }^2\right)},$$E5

while ${\sigma }^2$ denotes the variance of the misalignment.

4.2 All-optical cochlear implant

The CI implementations presented so far counterbalance either the RF scarcity that plagues the communications part of the system or the nerve stimulation limitations. To this end, the AOCI has been proposed as an architecture that converts the audio captured from the microphone into a light signal inside the external component for propagation to the cochlea [17] (Figure 4). This way, the AOCI counterbalances the aforementioned challenges and, at the same time, eliminates the need for an energy-consuming DSP unit in the implanted component. The AOCI not only builds upon the fruitful characteristics of the OWCI but also proposes breakthrough alterations such as the fact that it consists of only passive components, and thus, the implanted component has no power demands, which eliminates the requirement of complex power transfer policies and boosts energy efficiency. Furthermore, the AOCI utilizes optical neural stimulation, which is characterized by higher fidelity than electrical neural stimulation due to the lower spread of optical signals in human tissues. The technical advancements include the introduction of the AOCI architecture, its main building blocks, and the end-to-end system model. The AOCI takes into account channel, building block, and biological particularities [17]. Moreover, a novel tractable expression is derived for the instantaneous coupling efficiency in scenarios with misalignment fading. The feasibility of the proposed architecture is proven through the theoretical framework, which also evaluates its performance with regard to the power efficiency, the photon flux, and a plethora of design parameters that greatly influence the success or failure of the system.

Figure 4.

Illustration of the architecture of the AOCI. The all-optical nature of the AOCI resides in the combined utilization of optical wireless communications and optogenetics for stimulating the auditory nerve. Initially, the auditory neurons are sensitized to optical radiation with optogenetic techniques. Next, the sound captured from the external microphone is converted into an optical signal capable of stimulating the light-sensitive nerves, which is then forwarded to the cochlea.

Much like OWCI, the architecture of AOCI consists of the implanted and the external component, with the former located on the skull and the latter on the external surface of the skin. The external component captures the acoustic signal with a microphone, performs the necessary DSP, and converts it into the appropriate optical signal capable of generating the desired action potentials on the targeted spiral ganglion neurons. This signal is transmitted from the TX, which is a laser source, to the implanted component, where the guiding lens, the microelectromechanical device, the coupling lens, and the optical fiber ensure its delivery to the appropriate place in the cochlea. Specifically, the guiding lens guides the light toward the microelectromechanical device to maximize the power of the received optical signal. Afterward, the microelectromechanical acts as a mirror that mitigates the misalignment to a degree by steering the light beam to the center of the coupling lens in order to be coupled into the optical fiber. Finally, the latter delivers the light into specific points along the cochlea based on their spectral content.

4.2.1 Microelectromechanical device

Microelectromechanical devices have been the subject of much hype during the past decade due to their adaptability as well as low cost, low weight, and small size [80, 81, 82]. In the case of the AOCI, the microelectromechanical device is required in order to account for the individuality of each patient. In particular, the AOCI is required to adapt to the particularities of the patient, such as different skin thickness and color or slightly varied orientation of biological tissues, in order to ensure uninterrupted hearing restoration. Moreover, imperfections during the implantation process can cause slight variations to the final placement of the implant. To this end, the microelectromechanical device provides an externally operated light control system by enabling the steering of the optical beam toward the coupling lens. Finally, the microelectromechanical device adjusts its optical properties and, thus, steers the beam after receiving the appropriate electrical charge that can be applied during implantation, while in normal operation, the need for adjustment is eliminated, and therefore, the microelectromechanical device operates passively [83, 84].

The signal received by the guiding lens presented in (1) is forwarded to the microelectromechanical device, which introduces a collimation gain [85].

$$G_c=\frac{1}{\sqrt{{\left(1-d_{\mathit{\text{in}}}/f\right)}^2+z_0^2/f^2}}.$$E6

Therefore, the updated received signal at the output of the microelectromechanical device can be expressed as

$$y_2=G_c{h}_l{h}_{p}x+n.$$E7

4.2.2 Coupling lens

The coupling lens receives the optical beam from the microelectromechanical device and focuses it in the center of the optical fiber. The fact that incident light on the end of the optical fiber that arrives at a greater angle than the acceptable angle of the optical fiber is not coupled highlights the detrimental impact it plays on the maximum achievable coupling efficiency of the system. Moreover, the coupling efficiency is also affected by the dimensions of the coupling lens and the diameter of the optical fiber with its maximum value being in the order of 80% [86].

The coupling lens captures the optical signal that is reflected by the microelectromechanical device and couples it into the optical fiber. The signal that successfully enters the optical fiber can be written as

$$y_3=\eta G_c{h}_l{h}_{p}x+n$$E8

with the coupling efficiency given by

$$\eta ={\left(\frac{3.83\sqrt{2}D{\omega }_0}{1.22\mathit{\lambda F}}exp\left(-\frac{r^2}{{\omega }_0^2}\right){\mathrm{\Psi }}_2\left(1;2,1;-\frac{{3.83}^2{D}^2{\omega }_0^2}{{1.22}^2{\lambda }^2{F}^2},\frac{r^2}{{\omega }_0^2}\right)\right)}^2.$$E9

In Eq. (9), ${\omega }_0$, $F$, $D$, and $\rho$ denote the optical fiber mode field radius, the focal length, the focusing lens diameter, and the radial distance on the focal plane, respectively, while it becomes obvious that the achievable coupling efficiency is dependent on the optical fiber’s mode field radius, the coupling lens’s focal length and diameter, as well as the intensity of misalignment and the transmission wavelength.

4.2.3 Optical fiber

The optical fiber of the AOCI takes the place of the electrode array of the conventional CI. The incident optical signal must be delivered to specific locations alongside the cochlea in order to generate action potentials at the targeted spiral ganglion neurons that are responsible for the appropriate sound frequency. To achieve the required performance, the optical fiber proposed in the AOCI architecture propagates the optical signal through its single-mode core with a Gaussian beam profile in the output [87, 88]. Furthermore, despite the fact that state-of-the-art conventional CIs can be equipped with a maximum of 20 electrodes, due to the limited spatial resolution of electrical neural stimulation, the sound perceived by the patient has the fidelity of eight functional electrodes [40]. In addition, to achieve speech and music perception under suboptimal noise constraints, CI must house approximately 32 electrodes, which is also the goal of the AOCI [89, 90]. Therefore, tilted fiber Bragg gratings (FBGs) were introduced in the AOCI architecture that enable light delivery in various locations alongside the optical fiber [91, 92]. These FBGs are located in the core of the optical fiber, along the propagation direction, with a periodic variation of the refractive index. These components have low insertion loss, low complexity structures, and high wavelength selectivity. Specifically, tilted FBGs allow a limited number of wavelengths to penetrate them by filtering the incident optical signal based on its spectral content and, at the time, redirecting it based to their angle [93, 94].

When the optical signal travels through the optical fiber, it attenuates due to the curvature of the optical fiber and the existence of FBGs, and therefore, the emitted signal can be expressed as

$$y_4=\mathit{k\eta }{G}_c{h}_l{h}_{p}x+n,$$E10

where $k$ denotes the propagation efficiency, which is limited to $0.14\mathrm{dB}/{90}^{\circ }$ by the strong optical confinement of microfiber, even for increased bending radius or index values [95]. In addition, $k$ incorporates the signal attenuation due to the existence of FBGs, which has been proven to be in the order of 10% [96].

5. The road ahead

From the presented analysis, it is evident that, despite their extensive applications, electrical neural stimulation techniques suffer from insufficient coding of spectral information, low power efficiency, low stimulation precision, accuracy, and frequency, as well as questionable safety. To this end, promising optical neural stimulation methods that surpass these limitations have been proposed. In an effort to establish these methods, the scientific community has pushed toward proving their feasibility as well as theoretically modeling and augmenting them. The state of the art of optical neural stimulation techniques offers great promise toward realizing next-generation biomedical systems.

One of the main offerings of optical neural stimulation is the outstanding stimulation precision it offers compared to electrical neural stimulation. In more detail, the increased precision can be translated into higher customization of the produced neural activity in two respects. First, the increased stimulation frequency that comes with optical neural stimulation leads to higher accuracy of excitation due to the fact that action potentials are delivered faster to the target spiral ganglion neurons and, therefore, to the brain, thus limiting the time between sound acquisition and perception. Second, optical neural stimulation depends on the optical particularities of light sensitive opsins with each one being expressed in a specific type of cell. Therefore, this offers another layer of light selectivity that can be leveraged by optical neural stimulation techniques [97]. The combination of these two aspects equips optical neural stimulation with the necessary tools to achieve unprecedented performance not only in the field of hearing restoration but also in other biomedical application such as retinal implants that would utilize this advantage to provide higher perceived image fidelity.

Another aspect that boosts the performance of optical neural stimulation is the exceptional spectral coding of the information carried by the optical signal. On the contrary to electrical neural stimulation techniques that are characterized by wide current spread from the electrode contacts, optical radiation attenuates with a greater rate when it propagates inside human tissue, and therefore, the applied optical stimulations are more spatially confined than electrical ones. The importance of this phenomenon is highlighted even more by the fact that human sound perception requires at least 32 stimulation channels in order to recognize music or sound in noisy environments [89, 90]. As a result, the superior spectral coding of optical neural stimulation enables support for stimulation units that can house significantly more channels.

Contrary to previous detrimental improvements offered by optical neural stimulation methods, their performance in terms of power efficiency is comparable to the one of electrical neural stimulation. In more detail, optimization is required for optical neural stimulation stimulation policies in order to achieve similar power consumption as electrical neural stimulation [58]. Therefore, the optimization of optical neural stimulation techniques in terms of their power demands is one of the key requirements for their successful application in future biomedical applications. Similarly, the safety and ethical concerns of optical neural stimulation pose another controversial aspect. On the one hand, the optical power that is required for the reliable activation of light-sensitive spiral ganglion neurons is below the limits defined in various standardization protocols [98]; on the other hand, the modification of the targeted spiral ganglion neurons in order to acquire light sensitivity poses ethical concerns.

From a purely biological perspective, action potentials generated from electrical stimulation signals resemble the morphology and waveform of the membrane potential. As a result, these electrical signals are superimposed on each other and become almost indistinguishable, which hinders hearing restoration [58]. However, owing to its core functionality, optical neural stimulation triggers action potentials that differ significantly from membrane potential based on the stimulation protocol and the type of the excited cell. Specifically, not only the waveform of the generated action potential is affected by the amplitude and the duration of the stimulation, but also the instant release of ions when opsins are illuminated, which causes the membrane to react immediately. In addition, each opsin-cell-type combination is characterized by a distinct morphology of transmembrane potential and in conjunction with the wide variety of opsins available; they ensure the generation of a distinct action potential.

Finally, from an engineering point of view, the plethora of opsins that have been developed can highly impact the performance of optical neural stimulation biomedical applications. All future research in this field should take into careful consideration the selection of the applied opsin, as suboptimal ones may result in low stimulation precision and reliability, which, in turn, can determine whether the application is successful or not. The most important design choices include the compatibility with the target cell type, the amplitude and morphology of the resulting action potential, and the nature and the direction of the released ions.

6. Conclusions

In this chapter, we have provided a vision for hearing restoration from an engineering point of view that could serve as a guide in the research and development of the next-generation CIs. We suggest that the future of digital hearing restoration lies in the optical spectrum, both in terms of communication and stimulation techniques. We envisioned and explained potential architectures that enable the utilization of optical technologies in CIs. Finally, we introduced key features and performance indicators that could decide their success or failure.

Conflict of interest

The authors declare no conflict of interest.

Nomenclature

AOCI

all-optical cochlear implant

BER

bit error rate

ChR2

channelrhodopsin 2

CI

cochlear implant

DSP

digital signal processing

FBG

fiber Bragg grating

LED

light-emitting diode

MPE

maximum permissible exposure

OWCI

optical wireless cochlear implant

RF

radio frequency

SNR

signal-to-noise ratio