Spread of cochlear excitation during stimulation with pulsed infrared radiation: Inferior colliculus measurements

Abstract

Infrared neural stimulation (INS) has received considerable attention over the last few years. It provides an alternative method to artificially stimulate neurons without electrical current or the introduction of exogenous chromophores. One of the primary benefits of INS could be the improved spatial selectivity when compared with electrical stimulation. In the present study, we have evaluated the spatial selectivity of INS in the acutely damaged cochlea of guinea pigs and compared it to stimulation with acoustic tone pips in normal hearing animals. The radiation was delivered via a 200 μm-diameter optical fiber, which was inserted through a cochleostomy into the scala tympani of the basal cochlear turn. The stimulated section along the cochlear spiral ganglion was estimated from the neural responses recorded from the central nucleus of the inferior colliculus (ICC). ICC responses were recorded in response to cochlear INS using a multichannel penetrating electrode array. Spatial tuning curves were constructed from the responses. For INS, approximately 55% of the activation profiles showed a single maximum, ~22% had two maxima, and ~13% had multiple maxima. The remaining 10% of the profiles occurred at the limits of the electrode array and could not be classified. The majority of ICC spatial tuning curves indicated that the spread of activation evoked by optical stimuli is comparable to that produced by acoustic pips.

Introduction

In individuals with severe-to-profound hearing loss, cochlear implants (CIs) stimulate the remaining functional cochlear nerve fibers by applying electrical current directly into the cochlea. The devices are designed to take advantage of the tonotopic organization of the cochlea. They deliver electrical pulse trains derived from high-frequency sounds to the base of the cochlea and low-frequency signals to the apex of the cochlea [1-3]. In contemporary cochlear implants, the injected electric current spreads from an electrode or electrode pair along the scala tympani and across the cochlear turns. The spread of electric current in the cochlea has been examined in the cochlear implant user, in animals, in excised temporal bone preparations, and in computer models [4-12]; for a review see [13, 14]. All methods lead to a similar conclusion: stimulation of spatially discrete spiral ganglion cell populations is difficult with conventional methods and the interaction between electrodes could lead to activation of overlapping neural populations.

Current device development is aimed at cochlear implant designs that can stimulate smaller and independent populations of spiral ganglion neurons, thereby increasing the number of independent information channels delivered to the brain and improving user experience. Towards this goal, a novel approach has been suggested: pulsed infrared radiation (IR, ~1.86 μm) to stimulate spiral ganglion neurons [15]. This approach provides a fundamentally different interaction with tissue than the electric current. The radiation is largely confined to the optical path. This was initially determined by using tone-on-light masking [16] and by staining spiral ganglion cells for c-FOS, which is expressed in activated spiral ganglion cells [16, 17]. The results suggest that discrete stimulation of spiral ganglion cell populations can be accomplished using IR [17]. The previous approaches, however, had limitations. Tone-on-light experiments can only be conducted in hearing animals and c-FOS staining only worked for stimulation levels that were well above the stimulation threshold.

The aim of the present study was to provide direct measurements describing the selectivity of the optical stimulation at threshold levels and after cochlear damage with neomycin. For this purpose, acoustically evoked pure tone spatial tuning curves (STCs) were obtained from neural responses in the central nucleus of the inferior colliculus (ICC) in each guinea pig. STCs obtained with acoustic stimulation of cochleae with normal function were compared with ICC spatial tuning curves evoked by optical stimulation of the acutely damaged cochlea in the same animal. The ICC retains a tonotopic organization that correlates to the tonotopic organization of the cochlea. Thereby, a multichannel penetrating electrode inserted perpendicular to the frequency planes of the ICC can simultaneously measure neural responses resulting from a large area of the cochlea. The results suggest that the majority of optically evoked neural responses are spatially selective.

Methods

Pigmented guinea pigs (200-600 g) of either sex were used in the experiments. All procedures were carried out in accordance with the NIH Guide for the Care and Use of Laboratory Animals and were approved by the Animal Care and Use Committee of Northwestern University.

Animal surgery and electrode placement

Anesthesia

Anesthesia was induced by an initial intraperitoneal injection of Ketamine (44 mg/kg bodyweight) and Xylazine (5 mg/kg bodyweight). A tracheotomy was made and a plastic tube (1.9 mm outer diameter, 1.1 mm inner diameter, Zeus Inc., Orangeburg, SC) was secured into the trachea. The tube was connected to an anesthesia system (Hallowell EMC, Pittsfield, MA) including a vaporizer (VetEquip, Pleasanton, CA) to maintain anesthesia with isoflurane (1-3%). Depth of anesthesia was assessed with a paw withdrawal reflex and isoflurane concentration was adjusted accordingly.

Access to the cochlea

The cochlea was accessed by surgically opening the bulla. A c-shaped skin incision was made behind the left ear lobe and the cervicoauricular muscles were removed by blunt dissection. The cartilaginous outer ear canal was exposed and cut to insert an ear bar into the ear canal. A hollow ear bar on the left side and a solid ear bar on the right side were used to fix the head in a stereotactic head holder (Stoelting, Kiel, WI). The hollow bar allowed acoustic stimulation of the left ear, and the solid ear bar blocked the right ear canal, in addition to maintaining the head in a fixed position. The left bulla was exposed and opened approximately 2×3 mm with a motorized drill (World Precision Instruments, Sarasota, FL). The basal turn of the cochlea was identified and a cochleostomy was created with a 0.5 mm Buckingham footplate hand drill (Richards Manufacturing Co., Memphis, TN) approximately 0.5 mm from the bony rim of the round window. After the cochleostomy was made in the basal cochlear turn, an optical fiber (P200-5-VIS-NIR, Ocean Optics, Dunedin, FL) was inserted through the opening of the cochlear wall. For the present experiments, the fiber was 200 μm in core diameter, the numerical aperture was 0.22 and the acceptance angle was 25.4° in air. The optical fiber was mounted to a micromanipulator (MHW103, Narishige, Tokyo, Japan) to ensure consistent orientation during stimulation. Within the cochleostomy, the orientation of the fiber could be changed along the anteroposterior axis by approximately ±30° and along the dorsoventral axis by approximately ±45°. To measure compound action potentials (CAPs), a silver ball electrode was placed on the round window.

Placing the multi-channel ICC recording electrode

A silicon-substrate, thin-film multichannel penetrating electrode array (A1x16-5mm-100-177, NeuroNexus Technologies, Ann Arbor, MI) was placed in the right (contralateral) ICC. Each array had 16 recording sites (177 μm²/site) along a single shank at center-to-center intervals of 100 μm. To access the ICC, the right temporalis muscle was reflected, and an approximate 5×5-mm opening was made in the right parietal bone just dorsal to the parietal/temporal suture and just rostral to the tentorium. A small incision in the dura mater was made and the mutli-channel electrode array was advanced through the occipital cortex into the ICC using a 3D-micromanipulator (Stoelting, Kiel, WI). The latter was attached to the stereotactic head holder. The electrode array was inserted into the ICC on a dorsolateral to ventromedial trajectory at approximately 45° off the parasagittal plane in the coronal plane. Using this trajectory, the electrode array passed through the central nucleus of the ICC approximately orthogonal to its isofrequency laminae [18, 19]. After the initial placement of the distal tip of the electrode into the ICC, the electrode was advanced while an acoustic tone pip was presented to the left ear. Proper placement of the electrode was determined when neural responses from the distal contact of the array could be stimulated with a tone pip between 16-25 kHz. In some instances, the electrode was advanced several times into the ICC before the desired placement was achieved.

After placing the electrode array, the exposed skull and dura mater were covered and protected from dehydration with gauze sponges (Dukal Corporation, Hauppauge, NY) soaked with Ringer’s lactate solution. The ICC recording electrode array was placed, in most cases, prior to creating the cochleostomy because creating the cochleostomy could result in an elevation of the CAP thresholds. An elevation of CAP thresholds might distort the ICC electrode “mapping” and lead to a poor correlation of the ICC location to the best acoustic frequency.

Stimulus generation and calibration

Pure tone generation

Voltage commands for acoustic stimuli were generated using a computer I/O board (KPCI 3110, Keithley, Cleveland, OH) integrated into a PC and were used to drive a Beyer DT 770Pro headphone (Beyerdynamic, Farmingdale, NY). The speaker was coupled with a short, 3 mm-diameter plastic tube to the opening of the hollow ear bar. Acoustic stimuli were tone pips (12 or 20 ms duration, including a 1 ms rise/fall) with different carrier frequencies, which were presented at a rate of 4 Hz. The resulting sound level at the end of the hollow ear bar was measured with a 1/8-inch microphone (Bruel & Kjaer North America Inc., Norcross, GA). For the measurements, the microphone’s protective grid was flush at the opening of the ear bar.

Optical stimulation

Optical stimulation was achieved with an experimental, custom designed diode laser (Capella, Lockheed Martin Aculight Corp., Bothell, WA). For the present experiments, the wavelength was selected at 1.86 μm and the pulse duration at 100 μs. The laser was operated at 10 Hz repetition rate and was coupled to an optical fiber with a core diameter of 200 μm (see also above). The FWHM value for the optical spot measured at the tip of the optical fiber was 130 μm in diameter, with a Gaussian energy distribution. The radiant energy per pulse at the tip of the optical fiber was measured in air with the J50LP-1A energy sensor (Coherent, Santa Clara, CA) and was 0-127μJ/pulse. However, the energy measured at the tip of the optical fiber will not be the same as the energy delivered to the neurons. Fluids, soft tissue, and the modiolar bone absorb and scatter the radiation. The penetration depth of the radiation at 1.86 μm is about 700 μm, assuming primarily water absorption [e.g. 20]. Using post-experimental reconstructions of the stimulated cochleae, distances between the optical fiber and the spiral ganglion cells could be determined and were between 200 and 500 μm [21]. Since the incident energy decreases in water by 1/e for each 700 μm traveled along the optical path, it is a fair assumption that the energy at the spiral ganglion cells is about one third of the energy measured at the tip of the optical fiber.

Damaging the cochlea

After measuring the acoustic responses and verifying the placement of the ICC electrode, the cochlea was damaged by injecting an ototoxic drug into the scala tympani. With a small paper wick, some of the fluids in the scala tympani were removed under visual control. The fluid was then replaced by up to 50 μl of freshly prepared neomycin (20 mM in phosphate buffered saline at 38°C). The neomycin injection was done with a 100 μl Hamilton glass syringe. A filling tube (MicroFil MF34G, 70 mm long, 164 μm outer diameter, World Precision Instruments Inc., Sarasota, FL) attached to the syringe was inserted through the cochleostomy and the volume was manually flushed into the cochlea over ≤5 minutes. After the procedure, the CAP threshold elevation was measured. None of the animals showed a CAP for acoustic stimulus frequencies above 20 kHz following neomycin injection. About 50% of the animals had no acoustically-evoked responses between 15 and 20 kHz. Threshold elevations for frequencies between 8 and 15 kHz were at least 40 dB above baseline. At the stimulus frequencies below 8 kHz, CAP threshold elevations were at least 20 dB SPL.

Data acquisition

Compound action potentials

The round window electrode was connected to a differential amplifier (ISO-80, WPI, Sarasota, FL) with a high input-impedance (>10¹²Ω) and a gain of 60 dB. The responses were bandpass filtered (0.3 to 3 kHz) using the differential amplifier. The sampling rate was 250 kHz, and 32 responses to a pure tone stimulus were averaged for each data point. CAP thresholds were defined as sound levels required for a 30 μV peak-to-peak (P1-N1) response.

Inferior colliculus recordings

ICC recordings were made with the multi-channel electrode array and a Plexon data acquisition system (16-channel, Model MAP 2007-001, Plexon Inc, Dallas, TX). Neural activity was simultaneously recorded at 40 kHz sampling rate per channel, with a 16-bit analog/digital (A/D) input conversion. The bandpass filter range for the neural recordings was 0.1-8 kHz. Times at which spikes occurred were recorded for all 16 electrode contacts that were active. Figure 1 shows the neural activity recorded at 40-kHz sampling rate at one of the contacts during optical stimulation of the cochlea. The optical stimulus was 100 μs in duration, with radiant energy of 59.2 μJ/pulse. A field potential response along with single fiber activity can be seem. Following filtering to remove the field potential response, user-defined threshold determined the neural activity considered for further analysis (Fig. 1B).

Figure 1.

Shown is an example for recordings from one contact along the ICC electrode array. A shows the original trace (gray line) overlaid by the original data filtered by a low pass filter (cut off frequency 500 Hz, broken black line). The optical stimulus (100 μs pulse duration, radiant energy of 59.2 μJ/pulse) was presented at time 0. The original trace includes the evoked field potential at about 10 ms and the single fiber activity. To remove the field potential and low frequency components from the recording, the difference between the original and the filtered traces has been calculated and is shown in B. Circles indicate the times when action potentials were identified by a user-defined threshold. The times when the action potentials occurred were used for analysis.

Recordings were made while the cochlea was acoustically and optically stimulated. Acoustical stimuli were used to determine the best frequency for each electrode contact and to construct acoustically evoked STCs. Measurements were conducted at different acoustic stimulus frequencies between 1.0 and 32 kHz, at 3 steps per octave. For each frequency, the sound level was decreased from its maximum value over a range of 80 dB SPL in steps of 10 dB SPL. In three animals, measurements were repeated to demonstrate that the frequency contours for a particular ICC electrode placement did not significantly change over time. Responses to acoustic stimuli were recorded every hour for up to four hours. Differences among the trials were below the frequency resolution of the electrode array, ~0.3 octaves (data not shown). For optical stimulation, the energy was varied between 0 and 127 μJ/pulse at 100 μs pulse duration.

Data analysis and statistics

d-prime analysis

The number of action potentials at each of the 16 recording sites along the ICC electrode array was determined for different sound or optical energy levels. By comparing the number of action potentials for successive pairs of stimulus levels, the discrimination index, d′ (d-prime), could be calculated [22-24]. For the analysis, spike counts resulting from each of 20 stimulus presentations at one level and spike counts resulting from each of 20 stimulus presentation at the next higher level were compiled. An empirical receiver operating characteristic (ROC) curve was constructed. The ROC curve is a graphical presentation of the true positive versus false positive rate. The area under the curve was computed to express a z-score, which was then multiplied by √2 to derive the d′ value. In other words, the area under the curve is the probability that a classifier will rank a randomly chosen positive instance (rate did increase with increase in stimulus level or threshold was reached) higher than a randomly chosen negative one (rate did not increase with increase in stimulus level, threshold was not reached). It has been shown that the area under the ROC curve is closely related to the Mann–Whitney U test as well as the Wilcoxon test of ranks. The cumulative d′ was the accumulation of d′ values across successively increasing stimulus levels [25].

Spatial tuning curves (STCs)

STCs were obtained for pure tone and for optical stimuli. For the optical tuning curves, the stimulation site was kept constant by fixing the location of the optical fiber in the cochlea. The energy was varied while the neural activity was recorded at the 16 sites along the electrode array. In all cases, d′ analysis was performed for each contact. The results were then assigned to an n×m-matrix, where n corresponded to discrete sound levels or radiant energies and m to the different electrode contacts from 1 to 16. Using MATLAB, a contour plot was generated from the n×m-matrix. The threshold of stimulation was selected at the d′=1 iso-contour and the selectivity of the response was determined by measuring the width of the STC at the sound level corresponding to the tip of the d′=2 (Fig. 2). The value of d′=2 was selected as it corresponds to a sound level approximately 20 dB SPL above threshold, which was used in a previous paper to determine the width of the STCs for acoustical stimulation [18]. The width measures were then normalized to the total length of the multichannel electrode, which is 1500 μm. A similar approach was used for optical stimulation. After determining the threshold radiant energy for stimulation, the width of the STC at energy level corresponding to d′=2 at the primary contact was measured.

Figure 2.

Spatial tuning curve (STC) obtained with a pure tone acoustic stimulus at 3175 Hz. The color encodes the cumulative d′ value. The width of the STC is determined by measuring the width of the curve at a d′=2 (between the green arrows). The width of the STC is then normalized to the total length of the multi channel electrode array (1500 μm). The colorbar provides the cumulative d′ value.

Statistical evaluation

Average, standard deviation, median and quartile ranges were calculated for the data. Some of the data are shown in whisker plots. Mean and standard deviation were calculated for the widths and the STCs. For d′=2, the width for STCs obtained with optical and acoustical stimulation were compared. The comparison was made for all data and after sorting them into octave frequency ranges 1-2 kHz, 2-4 kHz, 4-8 kHz, 8-16 kHz, and 16-32 kHz. To determine whether differences between averages were significant, a Wilcoxon Rank test was performed if all data were used (optical versus acoustical) and a one-way analysis-of-variance (ANOVA) was done for the data sorted into their corresponding frequency ranges. If significant overall changes were found, pair-wise comparisons were made using a Tukey-honestly-significant-difference test with a 5% significance criterion.

Results

The results were obtained from 31 guinea pigs. STCs for acoustic stimulation were used to map the multi-channel electrode as described above. Figure 3 shows examples of acoustic STCs that were obtained from the ICC of one animal. For each given stimulus frequency, which is shown at the top of each plot, the neural activity at the 16 electrode contacts (y-axis), which was recorded for different stimulus levels (x-axis), was used to construct spatial tuning curves. As described in Methods, from the evoked rate, the cumulative d′ values were calculated. The colorbars show the cumulative d′ values. From the minima in the contour plots, the best frequencies at each electrode contact could be determined. The values for the best frequencies obtained from 31 guinea pigs are summarized in Figure 4. Figure 4A provides an example for a single electrode; the deepest electrode contact (contact 1) recorded neural activity from a region of the ICC that responded to acoustic stimulation at 25.4 kHz. The most superficial electrode (contact 16) corresponded to the ICC region that responded to ~1 kHz acoustic stimulation. The best frequency along the length of the electrode changed tonotopically and was consistent across animals (Fig. 4A and 4B). On average, the best frequencies increased from 2.0±1.6 kHz at the most superficial electrode contact to 18.5±3.0 kHz along the length of the multichannel electrode array (Fig. 4C). The values that are shown Figure 4C are the median, the 25^th and 75^th quartiles (boxes), and the total ranges (whiskers).

Figure 3.

Spatial tuning curves for pure tone stimuli from one guinea pig are shown. The deepest electrode contact recorded responses to ~25.4 kHz stimuli, whereas at electrode 11 responses to pure tone stimuli of 2 kHz were recorded. The colorbar provides the cumulative d′ values.

Figure 4.

Electrode-frequency mappings from 31 guinea pigs are shown. A presents an example of the best frequencies obtained at each electrode contact for acoustic stimulation as a function of depth into the ICC. Most experiments resulted in a frequency range represented on the electrode array between 2 and 18.5 kHz. B shows the best frequencies obtained at each electrode contact for acoustic stimulation as a function of depth into the ICC for all animals. C summarizes the frequencies determined for all guinea pigs used in this study. The number of the contact along the recording electrode can be converted into a measure of penetration depth into the inferior colliculus.

Inferior colliculus response profiles for INS

STCs in Figures 5 and 6 show the cumulative d′ values in plots of electrode contacts versus the optical energy (μJ). Among the 31 profiles recorded with INS, four general types of STCs were classified: (1) broad (responses with multiple maxima, N=4/31, Fig. 5A), (2) dual band (peaks at two distinct frequency bands, N=7/31, Fig. 5B), (3) focused (one best frequency, N=17/31, Fig. 5C) and (4) responses that were not fully in the range of the recording electrode (N=3/31, Fig. 5D). In the case shown in 5B, the best frequencies at electrode contacts 3 and 12 were 18 kHz and 4 kHz, respectively. In the case shown in 5C, the best frequency at contact 12 was 5.6 kHz. Overall, in the 17 “focused” profiles, optical stimulation targeted regions with best frequencies between 2-18 kHz.

Figure 5.

Four types of optically evoked ICC STCs are shown. A indicates a response curve showing multiple peaks located across much of the frequency range covered by the ICC electrode. This type of response was measured in 4/31 guinea pigs. B shows a response where two distinct peaks were seen (7/31 guinea pigs). C shows the most spatially restricted optical ICC map with only one peak (17/31 guinea pigs). D is representative of responses that could not entirely be measured with the current placement of the ICC electrode (3/31 guinea pigs). Colorbars in each panel provide the cumulative d′ value. In addition, the best acoustic frequencies at the 16 electrode contacts mapped prior to chemically damaging the cochlea are shown to the right of each panel. In determining the width of the response profiles, multiple peak profiles are treated as one broad profile as indicated in A by the black line.

Figure 6.

Examples for single contour ICC maps where INS targeted spiral ganglion population coding for distinct acoustic frequencies are shown in A-E. Each contour has been taken from a different animal. For the examples shown, the best frequencies for the spatial tuning curves (STCs) range from approximately 4 kHz (A) to 15 kHz (E). Contours obtained with the beam path perpendicular to the spiral ganglion are narrower (e.g. C, GP121-L2) when compared to the contours for which the beam path is more tangential to the spiral ganglion (e.g. E, GP074-L2). Colorbars provide the cumulative d′ value.

In 55% of the animals, ICC responses with a single peak were obtained. For each of the profiles, corresponding acoustical calibration was used to determine the best frequencies at each electrode contact. Figure 6 shows five different examples of STCs plotted versus the tonotopic best frequencies. For example, the profile shown in Figure 5C has been plotted versus corresponding acoustic frequencies in Figure 6B. In this case, the INS evoked response was primarily on electrode contact 12, which responded to acoustic frequency of ~4 kHz. Thus, the plots show examples of the best frequency for the tonotopic sections along the cochleae, which were targeted with INS.

The width of the STCs increased with increasing radiant energies and the d′ values Figure 7A shows the width of the STCs measured at different d′ values (filled versus open circles). For different values of d′ the corresponding INS-evoked STC widths were: 233 ± 190 μm (d′=1.5, N=28), 357 ± 206 μm (d′=2, N=28), 468 ± 188 (d′=2.5, N=20), and 596 ± 230 μm (d′=3, N=10). For further analysis at d′=2, the widths of the INS-evoked STCs were compared to those obtained with acoustic stimulation (Fig 7B and 7C). For pure tones, the value for d′=2 corresponded to sound levels which were ~20 dB SPL above threshold. The width of acoustic STCs at d′=2 was on average 383 μm ± 131 μm (N=61). All values above are given as mean ± standard deviation. For the calculation of the average width of the pure tone STCs, 61 representative profiles were used. Profiles were not used if they were not completely in the recording range of the electrode array (e.g., Fig. 3, top-left or Fig. 5D) and if the CAP-threshold curves were elevated.

Figure 7.

Shown are characteristics of optically and acoustically evoked STCs. A: The STC widths for d′ values between 1.5 and 3 at steps of 0.5 are shown. The width of the STCs increased for increasing d′ values. B: STC widths for optical (diamonds) and acoustical (circles) stimuli at a d′=2 are presented in the graph. On average the widths for optical stimulation were 357 ± 206 μm and 383 ± 131 μm for pure tone stimulation. Differences were statistically significant (p<0.05). C: The distribution of STCs widths for both stimuli at d′=2 from B are summarized. The distributions were similar for the optical and the pure tone stimuli.

The cumulative d′, which indicates a change in rate of firing of the neuron, increased with increasing optical energy. While in some neurons, the cumulative d′ increased over a factor of two change in energy, in others the cumulative d′ increased over a factor of ten (Fig. 8A). It is important to note that this set of data is biased because saturation of neural responses was not achieved in all experiments; rather the limit of the laser output was reached. The energy range over which the data were measured was limited by the maximum output energy of the laser of 127 μJ per pulse. The energy at stimulation threshold was 15.3 ± 11.3 μJ (Fig. 8B). No systematic trend of the thresholds with best frequencies could be observed.

Figure 8.

A shows the increase in d′ values with increasing radiant energy. Note that most of the neurons have cumulative d′ values below 4. The set of data is biased because the maximum cumulative d′ values are in part, limited by the energy range of the laser. For some neurons, the maximum response rate was not achieved at the maximum output of the laser. B: stimulation thresholds are shown. The average threshold was 15.3 μJ and did not change systematically in the examined frequency range.

Discussion

Spatial tuning curves (STCs)

For approximately 13% of the optically-evoked ICC STCs, multiple peaks were seen (e.g. Fig. 5A). It might be argued that multiple peaks are caused by structures in the optical path that block the radiation from stimulating the neurons (e.g. arrow in Fig. 9). In a separate series of experiments, reconstructions of the optical path have been made from serial sections of the guinea pig cochlea [21]. The optical paths obtained from the reconstructions were then compared with the ICC neural responses obtained in the same animal. Based on the results of the reconstructions, it is not clear whether bony structures in the optical path blocked the radiation and produced “shadows” in stimulation. Raster plots and post stimulus histograms examining the ICC neural activity revealed that the neural spike rate increased on the electrode contacts that produced “gaps” in the multi-peaked STCs, as well as at the electrode locations that showed a response on the STC. It is likely that these profiles, for most of the cases, actually represent one broad response area and that the “gaps” seen in the STCs are an artifact of the data analysis method. As such, multiple peak STCs were included in the data analysis and were treated as one broad response area (e.g. Fig. 5A).

Figure 9.

With the assumption that the spiral ganglion neurons are targeted by the optical radiation, the image shows the possible orientations of the optical fiber and the radiation beam along the length of the spiral ganglion. Shown is the cross section of a guinea pig cochlea with the nerve (N) in the center of the modiolus, and the neurons forming the spiral ganglion. The organ of Corti (OC) and the tectorial membrane (TM) can also be seen in the cross section. The beam path (indicated by the light gray bars) changes with the orientation of the optical fiber. Note, the scatter of the radiation by the fluids and by the bone has not been considered in the sketch. Orientation 1 results in a beam path that is tangential to the spiral ganglion, whereas orientation 2 is perpendicular to the spiral ganglion. Orientation 2 will irradiate a shorter length of spiral ganglion along the cochlea and can be expected to result in a narrower STC profile in the ICC. The arrow points towards a bony segment in Rosenthal’s canal that can reduce the radiant energy at the spiral ganglion neurons and may lead to multi-peak spatial tuning curves.

Approximately 55% of the STCs evoked by INS of the cochlea were narrow. The orientation of the optical fiber towards the spiral ganglion and the resulting optical path has a direct impact on the location of the ganglion being stimulated and the spread of stimulation (Fig. 9). For all the experiments, the access to the basal turn of the cochlea was at a similar location. However the site of stimulation varied along the cochlea since the orientation of the optical fiber was not always constant. Stimulation sites between 8 and 16 kHz lay directly in front of the optical fiber inserted through the basal turn cochleostomy and the optical path for these sites was perpendicular to the spiral ganglion (e.g. Orientation 2 in Fig. 9). To target lower or higher frequencies, the optical fiber had to be inserted at an angle into the scala tympani such that the stimulation site was within the optical path. In that case, the orientation of optical path would be more tangential to the spiral ganglion, likely resulting in a larger segment of the spiral ganglion being irradiated (e.g. Orientation 1 in Fig. 9). A broader STC would, therefore, be expected.

Single-peak optically-evoked STCs also showed best frequencies that do not correlate with locations directly in front of the optical fiber but rather with stimulation sites along the next cochlear turn. It is possible that locations closest to the optical fiber might not be stimulated, because the spot size is small and the optical path may not include spiral ganglion cells next to the tip of the optical fiber. A Gaussian beam profile was determined with the knife-edge measurement technique [26, 27]; the core-diameter of the optical fiber was 200 μm, and the angle of spread in Ringer’s Lactated solution was measured to be ~3°. Bone can scatter the radiation and the divergence angle of the beam after traversing an ex vivo section of modiolar bone in fluids was 26 degrees [26]. We have measured the placement of the tip of the fiber using microCT and observed it to be positioned between 300 to 700 μm away from the modiolar bone [28]. Combining the angle of spread and the distances of the tip of the optical fiber from the spiral ganglion neurons, the spot size at the spiral ganglion calculates to about 350 μm. Note that the assumption is that stimulation occurs at the spiral ganglion and not at the nerve fibers in the center of the modiolus.

When optically and acoustically evoked STCs were compared, the average width of the optically evoked STCs was 357 ± 206 μm and that of pure tone STCs was 383 ± 131 μm. The differences were statistically significant. To discuss why STCs obtained with optical sitmulation are expected to be smaller than those obtained with acoustic stimulation, we first assume that optical stimulation occurs at the spiral ganglion [21]. According to Ren [29], the length of the basilar membrane that vibrates during pure tone stimulation at moderate sound levels (~60 dB SPL) is about 600 μm in the gerbil. With a 1500 μm/octave slope of the place-frequency map [30], 600 μm corresponds to about 0.4 octaves. In the guinea pig, the slope of the place-frequency map is about 3200 μm/octave [31] and 0.4 octaves would correspond to a distance of 1280 μm along the cochlea. Considering that the spiral ganglion has only half the length of the basilar membrane, the corresponding distance would be 640 μm along the spiral ganglion. As shown above, the optical spot size at the spiral ganglion neurons with infrared stimulation can be estimated to 350 μm. This value is 1.8 times smaller than the 640 μm for acoustical stimulation. For acoustical stimulation, as described before, the width of the STC obtained from the ICC is ~383 μm. For INS the expected corresponding width can be estimated and is ~212 μm (383 μm/1.8). Figure 7C shows that about 4% of STCs obtained with optical stimuli are below 100 μm and about 21% of the STCs widths are below 200 μm, and 32% of the STCs widths are below 300 μm. The calculated values are similar to those expected for the beam path being perpendicular to the spiral ganglion. If the optical path interacted with the spiral ganglion to a degree that was not perfectly perpendicular (i.e. more tangential), the length of the stimulation site along the cochlea increases (Fig. 9).

Data for STCs obtained with pure tone stimuli and with electrical stimuli have been published for the guinea pig [18]. For pure tone stimuli the width of the STCs was on average 382 μm ± 144 μm, which compares well with the values from the present study. Visual placement of bipolar ball electrodes to stimulate the cochlea electrically resulted in STCs widths of 429 μm when radially placed and about 700 μm when longitudinally placed. STCs widths of 548 and 684 μm were reported for banded tripolar and banded bipolar electrodes [18]. For monopolar stimulation, the width of the STCs was larger than 1500 μm for both the banded and the ball electrode.

The optical penetration depth of water, the main absorber involved in INS, is wavelength dependent. For the present experiments the wavelength (1.86 μm) and consequently the penetration depth were kept constant. For radiation wavelengths between 1.83 and 1.90 μm, the water absorption curve is steep; i.e. small changes in wavelength result in large changes in optical penetration depth. The effect of changing penetration depth on neural stimulation/recruitment can be seen in the change in optically-evoked CAP amplitude [32]. The CAP amplitude increased with increasing penetration depth of the radiation, which was explained by an increasing number of neurons in the optical path being irradiated. Thus, it should be possible to employ small changes in wavelength to ensure that the desired cochlear structures are being irradiated.

Dynamic range

The normal cochlea can encode sound pressure levels over a range of about 120 dB SPL. For electrical stimulation and conventional intra-scalar stimulation paradigms, the dynamic range between the current amplitude at threshold and at maximum response is about 8-12 dB [33]. More recently, a penetrating nerve array has been proposed as an alternative for electrical stimulation of the auditory nerve. A larger dynamic range of stimulation (13-17 dB) has been reported with this technique [25]. The results of the present experiments indicate that the dynamic range for optical stimulation is greater than that for contemporary cochlear electrical stimulation. Moreover, the dynamic range for optical stimulation was limited in the present experiments by the stimulation threshold and the output power of the laser. In general, the stimulation threshold was at 15.3 μJ and the maximum energy per pulse (at 100 μs) from the laser was 127 μJ, which corresponds to a factor of 8.3 by which the energy could be increased. However, it is not clear whether electrical and optical values can be compared directly because optical energy is reported as a measure for INS, while current amplitude is used for electrical stimulation.

The more important question is whether optical stimulation indeed provides a greater number of discriminable level steps compared to cochlear electrical stimulation. This may be assessed by comparing each level with the next discriminable level step inducing a d′ change of 1. This approach has been applied by Middlebrooks and Snyder [25]. The number of stimulus levels between the cumulative d′=1 and d′=3 contours were measured at the electrode contact in the ICC that showed the lowest threshold. With those numbers, a “discrimination slope” was calculated. For electrical stimulation, this slope was expressed in units of d′ per decibel. For electrical stimulation with an intraneural electrode the slope was 0.73 d′, for a monopolar and bipolar electrode configuration the values were 1.98 and 1.94 d′, respectively [25]. A similar calculation was performed for the present results obtained with optical stimulation. When the slope is expressed in units of d′ per energy (μJ), it was on average 0.08 d′ (the average energy difference is 25.3 ± 21 μJ, N=10). Expressed in $\mathit{dB}=10\ast \log (\frac{{\mathit{energy}}_{d\prime =3}}{{\mathit{energy}}_{d\prime =1}})$ the slope was 0.42 d′ (the average dB value was 4.8 ± 2.3, N=10).

While the present results from acute studies demonstrate that optical stimulation may provide good dynamic range, we note that safety thresholds for long-term optical stimulation have not yet been defined. Photothermal interactions with the target or surrounding tissue at high radiant energies may alter tissue properties and reduce the dynamic range. Tolerance of optical stimulation at high radiant energies may also be a complicating factor. Long-term safety of optical stimulation and behavioral responses in awake animals need to be evaluated in the future.

Mechanism of optical stimulation

The mechanism of INS is still ambiguous. Consensus exists that INS occurs via a photothermal mechanism. The radiant energy is absorbed mostly by water and is subsequently converted into heat [34]. It is still unknown by which mechanism the temperature change is converted into an activation of neurons or other excitable cells, such as cardiomyocytes. INS responses in the vestibular system and cardiomyocytes cannot be explained by direct activation of afferent neurons, modulation of efferent synaptic inputs, or whole-organ temperature changes [35, 36]. Others have recently reported data for guinea pig ICC recordings in response to laser irradiation of the cochlea using a Q-switched, frequency-doubled Nd:YAG laser with a wavelength of 532 nm and a pulse duration of 10 ns [37, 38]. They have used laser parameters that resulted in stress confinement of the optical energy [39, 40]. Stress confinement results in a significant pressure wave, similar to an acoustical click. This constitutes a fundamentally different laser-tissue interaction than the photothermal mechanism of nerve stimulation employed in our experiments. Since the described method uses lasers to induce mechanical stimuli thereby activating the hair cells, they were not able to obtain responses in the deafened cochlea.

Laser induced pressure waves in water are well documented for experiments with high local absorption [22,23]. In a recent mansucript it has been reported that pulsed 1850 nm laser light used for neural stimulation also generates a measurable pressure [41]. For the laser’s maximum energy levels and with a 200 μm diameter optical fiber, the peak-to-peak sound pressure can reach 62 dB SPL in air. Since the cochlea is filled by endo- and perilymph, it is of interest to what extent laser induced sound waves exist when the absorbing volume is significantly decreased by immersion in water. Measurements in a swimming pool showed that radiant exposures of 0.35 J/cm² generated a pressure of 31 mPa in water. Referenced to 1μPa this corresponds to a value of 89.9 dB, while referenced to 20 μPa the above pressure would be 63.8 dB SPL. Considering an approximate gain of 26 dB through the middle ear, the pressure in scala tympani would correspond to a sound level at the ear canal of 37.8 dB SPL at the maximum power of the laser. A direct vibration of the basilar membrane and consequent hair cell stereocilia deflection is possible if hair cells are still present. The same measurements were performed with radiant exposures at stimulation threshold for infrared neural stimulation. The pressure in water was 0.35 mPa, which is 50.9 dB (re 1μPa) and 24.9 dB SPL. With a transfer function of 1:20 for the middle ear, 24.9 dB SPL would correspond to −1.2 dB at the ear canal. For the present experiments one has to keep in mind that pressure wave can contribute to the responses. However, they are likely small as has been outlined above. Moreover, our own experiments with chronic deaf animals that do not have hair cells show that optical stimulation can occur through an direct interaction between the optical radation and the auditory neurons [41].

Summary

We demonstrate that selective populations of neurons can be stimulated with optical radiation. The width of the response area in the inferior colliculus obtained with INS is similar to that obtained from acoustic tones. In some instances, the response area was observed to be narrower for optical stimulation when compared to responses obtained while stimulating with pure tones. Narrower stimulation areas suggest the possibility of an increased number of perceptual channels for cochlear implant user in devices based on optical stimulation. Experiments are underway to study the effect of simultaneous stimulation with two closely placed optical fibers to determine the optimum distance between the fibers which is necessary to obtain non-overlapping response areas in the ICC. Efficacy and short-term safety of INS was recently demonstrated in a cat model using continuous 200 Hz stimulation for up to 10 hours [42], but long-term safety will need to be demonstrated in a chronic animal model.