Chronic microstimulation in the feline ventral cochlear nucleus: physiologic and histologic effects

Hearing Research 149 (2000) 223^238 www.elsevier.com/locate/heares

Chronic microstimulation in the feline ventral cochlear nucleus: physiologic and histologic e¡ects D.B. McCreery *, T.G.H Yuen, L.A. Bullara Huntington Medical Research Institutes, Neurological Research Laboratory, 734 Fairmount Avenue, Pasadena, CA 91105, USA Received 3 March 2000; accepted 15 August 2000

Abstract This study was conducted to help to establish the feasibility of a multi-channel auditory prosthesis based on microstimulation within the human ventral cochlear nucleus, and to define the range of stimulus parameters that can be used safely with such a device. We chronically implanted activated iridium microelectrodes into the feline ventral cochlear nucleus and, beginning 80^250 days after implantation, they were pulsed for 7 h/day, on up to 21 successive days. The stimulus was charge-balanced pulses whose amplitude was modulated by a simulated human voice. The pulse rate (250 Hz/electrode) and the maximum pulse amplitude were selected as those that are likely to provide a patient with useful auditory percepts. The changes in neuronal responses during the multi-day stimulation regimens were partitioned into long-lasting, stimulation-induced depression of neuronal excitability (SIDNE), and shortacting neuronal refractivity (SANR). Both SIDNE and SANR were quantified from the changes in the growth functions of the evoked potentials recorded in the inferior colliculus. All of the stimulation regimens that we tested induced measurable SIDNE and SANR. The combined effect of SIDNE and the superimposed SANR is to depress the neuronal response near threshold, and thereby, to depress the population response over the entire amplitude range of the stimulus pulses. SIDNE and SANR may cause the greatest degradation of the performance of a clinical device at the low end of the amplitude range, and this may represent an inherent limitation of this type of spatially localized, high-rate neuronal stimulation. We determined sets of stimulus parameters which preserved most of the dynamic range of the neuronal response, when using either long (150 Ws/phase) or short (40 Ws/phase) stimulus pulses. Increasing the amplitude of the stimulus was relatively ineffective as a means of increasing the dynamic range of neuronal response, since the greater stimulus amplitude induced more SIDNE. All of the pulsed and unpulsed electrode sites were examined histologically, and no neuronal changes attributable to the stimulation were detected. There was some aggregation of glial cells immediately adjacent to some of the electrodes that were pulsed with the short-duration pulses, and at the highest current densities. ß 2000 Elsevier Science B.V. All rights reserved. Key words: Cat; Ventral cochlear nucleus; Auditory brainstem implant; Microelectrode ; Chronic stimulation

1. Introduction The auditory brainstem implant (ABI), which is based on an array of electrodes implanted over the human ventral cochlear nucleus (VCN), provides a workable auditory prosthesis for patients whose auditory nerve has been destroyed during resection of 8th nerve tumors, and who therefore cannot bene¢t from

* Corresponding author. E-mail: [email protected]

cochlear implants (Eisenberg et al., 1987; Shannon, 1989 ; Shannon and Otto, 1990; Shannon et al., 1993; Otto et al., 1997 ; Brackmann et al., 1993). However, the overall performance of patients with the surface electrode ABI is considerably poorer than is the average performance of patients with a multi-channel cochlear implant (Fishman et al., 1997; Parkin et al., 1985). This di¡erence may be due to the ABI's relatively poor access to the tonotopic organization of the lower auditory system, as compared to what is possible with the cochlear implants. We have been developing an auditory implant based on penetrating microelectrodes distributed across the tonotopic gradient of the VCN. This system may provide improved pitch discrimination

0378-5955 / 00 / $ ^ see front matter ß 2000 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 8 - 5 9 5 5 ( 0 0 ) 0 0 1 9 0 - 8

HEARES 3558 4-10-00

224

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

and improved speech perception in patients with little or no auditory nerve survival. Microstimulation permits excitation of a localized population of neurons, and thereby allows access to features of neuronal organization that are not accessible with larger (macro) electrodes (McCreery et al., 1998). However, the localization of the stimulus current to the immediate vicinity of the microelectrode also means that the neurons close to the microelectrodes are excited repeatedly. In contrast, with normal sensory inputs, or when stimulating with macroelectrodes, the neuronal activity would be distributed over a much larger population. One consequence of this persistent, focused stressor is a long-acting depression of neuronal excitability. A few hours of microstimulation in the VCN will cause prolonged stimulation-induced depression of neuronal excitability (SIDNE) if the rate of stimulation and/or the pulsing rate are high (McCreery et al., 1997). Furthermore, histologically detectable tissue injury can occur in the VCN after only a few hours of stimulation (with charge-balanced current pulses) at a rate of 500 Hz, and when the stimulus charge per phase continually exceeds 3 nC/phase (McCreery et al., 1992, 1994). Analyses of the cytoarchitecture of the human cochlear nuclei have shown that the same cell types are present in the feline and human nuclei (Moore and Osen, 1979; Moore, 1987). In our previous studies of the safety of microstimulation in the feline PVCN, most of our stimulation protocols spanned only a single day. However, data from one animal did indicate that longer regimens of intranuclear microstimulation could be safe, and would not induce excessive SIDNE, if the stimulation was not continuous. We therefore performed experiments in which iridium microelectrodes implanted chronically in the feline PVCN were pulsed for 7 h/day, on 10^21 successive days, in order to determine the physiologic and histologic consequences of these prolonged regimens of stimulation. We conducted stimulation protocols in which the duration of the charge-balanced stimulus pulses was either 150 Ws/phase or 40 Ws/phase. Either protocol may ¢nd use in a clinical device, depending upon the capabilities of the stimulator to which the microelectrodes are attached. The longer pulses are most compatible with the slow kinetics of charge injection across the metal^electrolyte interface of the activated iridium stimulating electrodes (Robblee et al., 1983), and should minimize the risk of any dissolution of the iridium electrodes, over the lifetime of the patient. However, the short pulses are most compatible with most of the extant clinical stimulators, which were designed to pulse (larger) electrodes over a greater range of amplitudes.

Fig. 1. A scanning electron micrograph of the tip of a stimulating microelectrode. The insulation (darker material) has been ablated by an erbium laser.

2. Methods 2.1. Fabrication of stimulating microelectrodes Iridium stimulating microelectrodes are fabricated from lengths of pure iridium wire, 70 Wm in diameter. A Te£on-insulated lead wire is welded to one end of the iridium shaft, and the other end is shaped to a conical taper, by electrolytic etching. The microelectrodes have relatively blunt tips (a radius of curvature of approximately 6 Wm) to reduce tissue injury during insertion into the brain. The entire shaft and wire junction then is coated with three thin layers of Epoxylite 6001-50 heat-cured electrode varnish. The insulation is removed from the tip by dielectric destruction or with an erbium laser, leaving an exposed geometric surface area of 1000 þ 200 Wm2 (Fig. 1). The individual electrodes are assembled into an integrated array of four microelectrodes spaced approximately 400 Wm apart. The iridium electrodes are `activated' to increase their charge capacities (Robblee et al., 1983), cleaned with a detergent, soaked in deionized water for 24 h, and sterilized with ethylene oxide.

HEARES 3558 4-10-00

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

225

2.2. Implantation of stimulating and recording electrodes Young adult cats were anesthetized with pentothal sodium, with transition to a mixture of nitrous oxide and halothane. Implantation of the electrodes is conducted using aseptic surgical technique. A stainless steel recording electrode was implanted by stereotaxis into the right inferior colliculus (IC), and a reference electrode was implanted about 2 mm dorsal to the colliculus. A small craniectomy is made over the cerebellum, through which the array of iridium stimulating electrodes is implanted into the left posteroventral cochlear nucleus (PVCN). The PVCN and the posterior position of the anteroventral cochlear nucleus are potentially useful locations for an intranuclear auditory prosthesis. The multipolar (stellate) neurons of this region project to the nuclei of the lateral lemniscus and to the IC, and therefore, appear to be one of the cell populations that mediates monaural hearing (Cant, 1982; Brunso-Bechtold et al., 1981 ; Warr, 1972). 2.3. Stimulation protocols and data acquisition The regimens of prolonged stimulation were initiated 80^250 days after implantation of the electrodes. Three of the four microelectrode were pulsed for 7 h/day, on up to 21 consecutive days. SIDNE during microstimulation in the feline VCN is strongly dependent upon the pulsing rate (McCreery et al., 1997), and therefore in a clinical prosthesis, we should not stimulate at a higher rate than is absolutely necessary. Although some studies of cochlear implant patients with multi channel systems have indicated improved performance at high stimulation rates (Brill et al., 1997; Wilson et al., 1995) other studies have shown little di¡erence in the performance in cochlear implant patients using the continuous interleaved sampling strategy, as the pulse rate was varied from 200 to 4000 Hz/electrode (Shannon et al., 1995). Based on these ¢ndings, and the importance of minimizing SIDNE, we selected a pulsing rate of 250 Hz/electrode. For the prolonged stimulation regimen, we have simulated an acoustic environment based on a computergenerated arti¢cial voice that was speci¢ed and provided by the International Telegraphic and Telephony Consultive Convention, for the purpose of testing telecommunication equipment (CCITT, 1988). The CCITT arti¢cial voice reproduces many of the characteristics of real speech, including the long-term average spectrum, the short-term spectrum, the instantaneous amplitude distribution, the voiced and unvoiced structure of speech, and the syllabic envelope. The arti¢cial voice signal was passed through a full wave recti¢er and then underwent logarithmic amplitude compression, before being sent through an appropriate anti-aliasing ¢l-

Fig. 2. The distribution of the amplitudes of the stimulus current pulses generated from the CCITT arti¢cial voice signal, after full wave recti¢cation, low-pass ¢ltering over the range of 0^100 Hz and logarithmic amplitude compression. The distribution of pulse amplitudes has been scaled and shifted so as to range from 6 to 20 WA.

ter. The amplitude of the signal from the ¢lter then sets the amplitude of the charge-balanced stimulus pulses, which were delivered to each electrode at 250 Hz/electrode, in an interleaved manner. The range of pulse amplitudes was shifted so that acoustic silence is represented by a pulse amplitude that is close to the threshold of the response recorded in the central nucleus of the IC while stimulating in the PVCN. The threshold of the evoked response is typically 6 þ 1 WA when the stimulus pulse duration is 150 Ws/phase, or 14 þ 2 WA, when the pulse duration is 40 Ws/phase. Fig. 2 shows the amplitude distribution of the current pulses generated from the logarithmically compressed CCITT arti¢cial voice signal, when the pulses are scaled to range between 6 and 20 WA. The arti¢cial voice signal was presented for 15 s, followed by 15 s in which the stimulus amplitude was held near the threshold of the evoked response (6 or 14 WA, as noted above). Each of the pulsed electrodes received the same stimulus, but with the stimulus pulses interleaved across the array. This 50% duty cycle is intended to simulate a moderately noisy acoustic environment. The stimulation and data acquisition were conducted using a two-way radiotelemetry stimulation and data acquisition system, and the cats are able to move about freely in a large Lucite cage. This telemetry system and its companion software allow continuous monitoring of the voltage waveform across the stimulating microelectrodes, and of the compound evoked potential induced in the IC by the stimulating microelectrodes.

HEARES 3558 4-10-00

226

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

Fig. 3. An AER induced by stimulating in a cat's PVCN and recorded in the central nucleus of the contralateral IC. The amplitude of the cathodic-¢rst stimulus pulse was 16 WA and the pulse duration was 150 Ws/phase. The response has two prominent components. The RGFs were generated from the AER's ¢rst component, whose amplitude was measured from the peak of the ¢rst positivity to the trough of the subsequent negativity.

Before and after each daily session of stimulation, 1024^4096 responses evoked from each of the microelectrodes in the PVCN were recorded in the IC. These responses were averaged to obtain averaged evoked responses (AERs). The amplitude of the ¢rst and second components of the AER was measured from the peak of the positivity to the trough of the subsequent negativity (Fig. 2). Because of its short (V1 ms) latency after the stimulus, the ¢rst component is assumed to represent neuronal activity evoked directly in the neurons projecting from the PVCN to the IC, while the second component may represent neuronal activity that is evoked transsynaptically. All of the response growth functions (RGFs) described in this report were generated from the ¢rst component. In order to determine an appropriate range of stimulus pulse amplitudes, in one animal we compared the AER induced by current pulses delivered through a microelectrode implanted chronically in the left PVCN, and the AER evoked by acoustic stimuli (compression^rarefaction clicks at 50/s) delivered to the left ear by a speaker-phone positioned 10 cm lateral to the pinna. The cat was lightly anesthetized (with ketamine), the head stabilized, and the right external ear canal occluded with closed-cell foam rubber. The amplitude of the acoustic stimuli was calibrated at the cat's pinna using a Bruel and Kjaer 4176 omnidirectional microphone and a B and K 2235 sound level meter. The upper trace in Fig. 4A shows the AER evoked in the IC by clicks with a mean amplitude of 1.5 dynes/cm2 , or

approximately 55 dB above the sound pressure reference level of 0.0002 dynes/cm2 . The lower trace shows the response evoked by stimulating in the PVCN with cathodic-¢rst, biphasic current pulses, 20 WA in amplitude and 150 Ws/phase in duration. The ¢rst component of the electrically evoked and acoustically evoked responses is comparable in amplitude and duration, but the electrical stimulus evoked a larger late response. As noted above, the ¢rst component of the AER probably represents activity evoked directly in the cochlear nucleus neurons that project to the IC, while the second response probably represents neuronal activity that is evoked transsynaptically. It is not clear why the electrical microstimulation induced the larger second response, but it may be related to the very di¡erent spatial pattern of neuronal activity evoked by the electrical and acoustic stimuli. The abscissa of Fig. 4B is the amplitude of the (electrical) current pulses, and the ordinate is the amplitude of the acoustic clicks required to induce an AER whose ¢rst component is of the amplitude listed alongside of the curve. Thus, 15 WA and 0.45 dynes/cm2 will both evoke early components of approximately the same width, and each with an amplitude of 8 WV. This is an indication that the two stimulus modalities were exciting approximately the same number of neurons, although almost certainly not the identical populations (the relation between a population response such as the AER, and the underlying neural activity, is addressed in Section 4). Our primary objective is to develop a clinical ABI which will allow improved comprehension of speech. At a distance of 1 m, the mean SPL of the principal author's voice is approximately 55 dB. Therefore, when the logarithmically compressed arti¢cial voice modulates the stimulus pulses over the range of 6^20 WA, the neural response is similar to that evoked by an acoustic signal ranging from 0 to 55 dB at the cat's pinna. In Section 3, we describe a procedure for re-scaling the pulse amplitudes when the pulse duration is other than 150 Ws/phase. The RGFs, which represent the recruitment of the neural elements surrounding the microelectrode, were generated for each stimulating microelectrode, by plotting the amplitude of the ¢rst component of each of the AERs evoked from the microelectrode, against the amplitude of the `probe' stimulus that evoked the AER. The growth functions were generated before, and also immediately after the 7-h sessions of stimulation. The frequency of the `probe' stimulus was 50 Hz, which is much lower than that of the 250 Hz used during the 7-h stimulation session. Since these growth functions were not acquired during the 7-h session, and not contemporaneously with the arti¢cial voice signal, we refer to them as `non-embedded RGFs'. In addition, a limited number of `embedded' RGFs were acquired during the ¢rst or last hour of some of the 7-h sessions of stimu-

HEARES 3558 4-10-00

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

Fig. 4. A: Acoustically evoked and electrically evoked AERs recorded in a cat's IC. The acoustic response was elicited by a train of compression^rarefaction clicks with a mean sound pressure level of 1.5 dynes/cm2 at the pinna of the cat's left ear. The electrical response was evoked by microstimulation in the cat's left PVCN using 20 WA, 150 Ws/phase current pulses. B: The correspondence between the amplitude of the electrically evoked and acoustically evoked AERs. The numbers adjacent to the curve indicate the amplitude of the ¢rst component of the AER, as depicted in A. The abscissa and ordinate indicate the amplitude of the acoustic clicks or of the electrical pulses required to evoke an AER of the amplitude indicated by the values printed adjacent to the curve.

lation with the arti¢cial voice signal, at the full stimulation rate of 250 Hz/electrode. When the appropriate pulse amplitude (e.g. 6 þ 0.5 WA) is generated in accordance with the logarithmically compressed arti¢cial voice, subsequent stimulus pulses are suspended for

227

8 ms, which is su¤cient for the ¢rst and second components of the response evoked from the kth electrode to be recorded in the IC, without interference from the next stimulus artifact generated by the kth or other microelectrodes. After 8 ms, pulsing at 250 Hz/electrode resumes, and the computer waits 200 ms before starting its surveillance for the next 6 þ 0.5 WA pulse. When detected, another response evoked from the kth electrode is recorded. This is repeated until the required number of responses from the kth electrode has been collected and averaged. Typically, 1024 or 2048 responses are summated (averaged) to generate an `active' AER. Next, the process is repeated but now, the 6 þ 0.5 WA pulse that would have been delivered to the kth electrode is replaced with a pulse amplitude of 0 (the replacement is done only when the response is to be recorded, so fewer than 1% of the stimulus pulses are missing). This `null' response is subtracted from the `active' AER to eliminate any portion of the response that was evoked from the other microelectrodes. The entire process is then repeated for a pulse amplitude of 8 þ 0.5 WA, 10 þ 0.5 WA, etc., until the complete growth function for that microelectrode has been generated. At least 60 min/microelectrode are required to generate the embedded RGF. The non-embedded RGFs reveal depression of neuronal excitability that persists after the end of the 7 h of high-rate stimulation with the arti¢cial voice signal. This type of persisting SIDNE may persist for many days (McCreery et al., 1997). The occurrence of marked SIDNE may identify stimulation protocols that place signi¢cant stress on the neurons of the VCN. The embedded growth functions reveal how the regimen of prolonged stimulation a¡ects these neurons' responses to the actual arti¢cial voice signal. In particular, they reveal short-acting neuronal refractivity (dubbed `SANR'), which is not detectable in the non-embedded responses. The non-embedded and embedded responses, and the histologic evaluation of the implant sites, together provide a picture of the safety and e¤cacy of the stimulation regimen. Within 15 min after the end of the last day of stimulation, the cats were deeply anesthetized with pentobarbital and perfused through the aorta with 1/2 strength Karnovsky's ¢xative (2.5% glutaraldehyde, 2% paraformaldehyde and 0.1 M sodium cacodylate bu¡er). The cochlear nucleus was resected, embedded in para¤n, sectioned serially in the frontal plane (approximately parallel to the shafts of the stimulating microelectrodes) at a thickness of 8 Wm, and stained with Nissl stain or with hematoxylin and eosin. The contralateral IC also was processed and sectioned, to con¢rm the location of the recording electrode.

HEARES 3558 4-10-00

228

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

3. Results 3.1. SIDNE, SANR and the e¡ect of stimulus pulse rate and pulse amplitude Fig. 5A illustrates the strong dependence of both the SIDNE and the SANR upon the stimulus pulsing rate. Three electrodes in this animal's PVCN were pulsed for 7 h/day on 12 successive days, with the stimulus pulse amplitude modulated according to the logarithmically compressed arti¢cial voice signal. During the ¢rst 9 days of the stimulation regimen, the pulsing rate was 100 Hz/electrode. The non-embedded RGFs, acquired before the start of the ¢rst day of stimulation (prestimulus growth function), the non-embedded growth function acquired after the 9th day, and the embedded growth function acquired late in the 9th day, were very similar. On the 10th day, the pulse rate was increased to 250 Hz/electrode (Fig. 5B). The non-embedded growth function acquired just after the 7-h session was shifted well to the right of the prestimulus non-embedded growth function, a manifestation of SIDNE. Also, the embedded growth function was shifted to the right of the non-embedded growth function acquired after the end of the 7-h sessions (a manifestation of SANR). The stimulation-induced changes in the growth functions depicted in Fig. 5B are typical for these stimulus parameters (pulse rate of 250 Hz, stimulus pulses modulated between 6 and 20 WA by the arti¢cial voice signal). There was relatively little change in the absolute threshold of the evoked response after the 7 h of stimulation, but the slope of the growth function near threshold had decreased markedly. At higher pulse amplitudes, the slopes of the pre- and post-stimulus and

growth functions were similar. This implies that the prolonged stimulation primarily a¡ected the excitability of the neurons that were close to the stimulating microelectrode (assuming that the neurons close to the electrodes gave rise to the evoked response near threshold). SIDNE of the magnitude depicted in Fig. 5B is reversible over an interval of 5^20 days after cessation of stimulation (McCreery et al., 1997). The stimulation-induced changes depicted in Fig. 5B are di¤cult to quantify as a single scalar, in part because the growth functions are nonlinear and become more so during the prolonged stimulation. The separation of the growth functions at the maximum pulse amplitude (20 WA in this case) provides an explicit, albeit rather crude, quantitation of the e¡ect of the stimulation on the RGFs. When the shift is downwards,

C Fig. 5. The e¡ect of stimulus pulsing rate during the 7-h daily sessions, on the non-embedded and embedded RGFs of the ¢rst component of the AER evoked in a cat's IC by microstimulation in the contralateral PVCN. The `non-embedded' growth functions were generated from AERs elicited by pulsing with a low rate (50 Hz) probe stimulus, before and after each of the 7-h daily sessions. The embedded RGFs were generated at the full pulsing rate of 250 Hz, near the end of each daily session, using pulses whose amplitude was modulated by the arti¢cial voice signal. The range of stimulus pulse amplitudes generated from the logarithmically compressed arti¢cial voice signal and applied throughout the 7-h daily sessions, is indicated above the abscissa. A: E¡ects of stimulating for 7 h/day with the arti¢cial voice, at a pulsing rate of 100 Hz. The growth functions remained very close to the prestimulus (control) non-embedded growth function. B: E¡ects of stimulating at 250 Hz/electrode. The non-embedded growth functions became shifted below and to right of the pre-stimulus non-embedded function and the embedded function acquired during the last hour of the third session was shifted to the right and below the non-embedded function acquired immediately after that session. Also shown is the method of computing the iSIDNE, iSANR, and iTOTAL.

HEARES 3558 4-10-00

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

229

Fig. 6. A: The iSIDNE for 16 microelectrodes from 12 cats, computed as illustrated in Fig. 5B, at the end of the ¢rst day and the last day of stimulation regimens, which spanned 15^21 days. The indices are displayed as a percent change, relative to the pre-stimulus RGF. During the 7-h daily stimulations, the stimulus pulses were modulated between 6 and 20 WA by the arti¢cial voice signal (12 microelectrodes) or between 6 and 32 WA (four microelectrodes). Data from individual microelectrodes are shown as open symbols. Filled symbols represent the means for the two ranges of stimulus pulse amplitudes. The error bars are the standard deviations of the two groups. B: The iSANR for eight of the same microelectrodes, shown as the percent change from the non-embedded RGF acquired at the end of the daily session. C: iTOTAL for these eight microelectrodes, as percent changes from the prestimulus non-embedded RGF.

as is nearly always the case, the separation of the RGFs is an index of SIDNE. We designate this as the `index of SIDNE' (iSIDNE) to distinguish it from the more general phenomenon which a¡ects the entire growth function. We chose this particular measure of SIDNE because the logarithmic compression of arti¢cial voice signal causes the pulse amplitude to be clustered in the upper part of its range (Fig. 2).

Fig. 6A shows the iSIDNE from 16 microelectrodes from 12 cats, measured at the end of the ¢rst and last day of the stimulation regimen. The electrodes had been implanted 80^250 days prior to the start of the stimulation. These 16 electrodes evoked response in the contralateral IC that had a distinct early component, as depicted in Fig. 3, from which the RGFs could be generated. Twelve of the electrodes were pulsed over the

HEARES 3558 4-10-00

230

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

range of 6^20 WA by the arti¢cial voice signal, and four were pulsed over the range of 6^32 WA. We selected this upper limit because nearly all of our cats would tolerate stimulation through three or four of the microelectrodes

at 32 WA, without exhibiting fear or startle responses. During the 7-h daily stimulation sessions, the stimulus pulse duration was 150 Ws/phase and the pulsing rate was 250 Hz. When the stimulus pulses generated from

HEARES 3558 4-10-00

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

231

Fig. 7. A: A histologic section, nearly parallel to the lower track and the tip site (T) of a microelectrode implanted chronically in a cat's PVCN. The microelectrode was pulsed for 7 h/day on 19 consecutive days. The stimulus pulse duration was 150 Ws/phase, and the arti¢cial voice signal modulated the stimulus amplitude between 6 and 20 WA (see text for other details of the stimulus regimen). The track and tip site are surrounded by a compact gliotic sheath and the surrounding neurons (N) and neuropil appear normal and healthy. B: The tip site (T) of another microelectrode in the PVCN of another cat. The stimulation regimen spanned 15 days, but otherwise was as described above. The neurons (N) and neuropil surrounding the tip site appear normal. C: The site of the tip (T) of a microelectrode from another cat in which the stimulus pulses were modulated between 6 and 32 WA by the arti¢cial voice signal, during the 15-day stimulation regimen. Neurons (N) and neuropil ventral to the tip are somewhat £attened but are otherwise indistinguishable from those surrounding the unpulsed microelectrode shown in D. Nissl stain (toluidine blue). Bar = 50 Wm. 6

the arti¢cial voice signal ranged from 6 to 20 WA, most (10 of 12) of the iSIDNE were less than 20% at the end of the ¢rst day of stimulation. The iSIDNE subsequently changed little throughout the stimulation regimens, which spanned 15^21 days. In contrast, when the pulse amplitudes ranged from 6 to 32 WA, all of the iSIDNE were greater than 35% at the end of the ¢rst day of stimulation. Although in most cases the SIDNE did not increase over the remainder of the 15-day regimens, it e¡ectively negated the initially larger response evoked by the greater pulse amplitudes. The di¡erence in the iSIDNE when the arti¢cial voice signal modulated the stimulus between 6 and 20 WA, vs. 6^32 WA, is statistically signi¢cant (P 6 0.001, two-sided t-test with unequal n). In spite of the small n, the combined e¡ect of the SIDNE and SANR also was signi¢cantly di¡erent for the two protocols (P 6 0.01). For both ranges of the arti¢cial voice signal, the SIDNE was not signi¢cantly di¡erent between the end of the ¢rst and last day of the regimen (P s 0.5 for both protocols). Fig. 6B shows the index of SANR (iSANR), as computed from the data from eight of the microelectrodes depicted in Fig. 6A. The index was computed as illustrated in Fig. 5B, from the separation between the embedded growth functions acquired during the last 90 min of the 7-h stimulation session and the non-embedded growth functions acquired immediately after the session. When the stimulus pulse amplitudes ranged from 6 to 20 WA, the iSANR were 25% or less, but reached 50% when the stimulus pulses ranged from 6 to 32 WA. Fig. 6C shows the index for the combined e¡ect of SIDNE and SANR (`iTOTAL'), which was computed as illustrated in Fig. 5B. The results depicted in Fig. 6 illustrate that prolonged stimulation in the PVCN, at a rate of 250 Hz/electrode, in most cases produced moderate SIDNE and SANR when the amplitude of the 150 Ws stimulus pulses did not exceed 20 WA. Fig. 7A shows a histologic section cut obliquely through the lower track and tip site of one of the pulsed microelectrodes from Cat CN122. The microelectrodes were pulsed for 7 h/day on 19 successive days, wherein the 150 Ws/phase current pulses were modulated over the range of 6^20 WA by the arti¢cial voice signal. The microelectrodes had been implanted for 112 days

at the time of killing. The electrode's tip was in the ventral part of the PVCN, and is surrounded by a compact gliotic sheath, approximately 10 Wm in thickness. The surrounding neuropil is compact and appears to be healthy, and there are no microvacuolations of myelinated axons that are a cardinal feature of stimulationinduced tissue injury in the PVCN (McCreery et al., 1992, 1994). There is no in¢ltration of in£ammatory cells. The multipolar and spherical neurons, some with-

Fig. 8. A histologic section through the lower track (TR) and tip site (T) of a microelectrode which penetrated completely through the ventral surface of the PVCN. The arti¢cial voice signal modulated the 150 Ws current pulses between 30 and 50 WA, at 250 Hz, for 7 h on 10 consecutive days. The stimulus range was shifted in order to compensate for the elevated threshold of the RGF, which resulted from the electrode tip being slightly outside of the PVCN. The mechanical distortion of the electrode track is an artifact which occurred during handling of the histologic sections. The neurons and neuropil in the PVCN immediately dorsal to the tip site appear to be normal, in spite of the high stimulus current. Nissl stain. Bar = 100 Wm.

HEARES 3558 4-10-00

232

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

in 50 Wm of the tip, appear normal, with no indication of chromatolysis or other injury. Fig. 7B shows a histologic section through the tip site of a microelectrode implanted for 165 days in the PVCN of another cat (CN108) in which the microelectrodes were pulsed for 7 h/day on 15 consecutive days, with the stimulus pulses modulated between 6 and 20 WA. The track and tip site are surrounded by a compact glial sheath approximately 15^20 Wm in thickness. There is no in¢ltration of in£ammatory cells and the nearby neurons and neuropil appear quite normal. Fig. 7C shows the site of the tip of one of the pulsed electrodes from animal CN111, in which the arti¢cial voice signal modulated the 150-Ws pulses between of 6 and 32 WA for 7 h/day on 15 consecutive days. The tip site is in the rostro-medial part of the PVCN and the multipolar neurons close to the tip appear quite normal. Some of the neurons do appear to be somewhat £attened, and the neuropil in the vicinity of the electrode tip appears to have been displaced by the rather blunttipped microelectrode. This phenomenon occurred at many of the tip sites, and its long persistence after implantation (in this case, 255 days) is curious. However, this pulsed tip site appears essentially indistinguishable from the site of the unpulsed electrode (Fig. 7D). There were no histologic changes that could be attributed to the electrical stimulation adjacent to any of the microelectrodes that were pulsed on 3^21 consecutive days, using stimulus pulses 150 Ws in duration, and in which the stimulus pulse amplitude reached 32 WA. If the electrical threshold is initially elevated, it may be feasible to transpose the entire range of the stimulus. Fig. 8 shows a histologic section through the lower part of the track of an electrode that had actually penetrated through the ventral surface of the cochlear nucleus, and the threshold of the non-embedded response was nearly 30 WA. This would be analogous to the situation in which an electrode passed through the human cochlear nucleus and came to reside in the inferior cerebellar peduncle. The electrode was pulsed for 7 h/day on 10 successive days, at 250 Hz and the 150-Ws stimulus pulses were modulated over the range of 30^50 WA by the arti¢cial voice signal. These parameters produced SIDNE by the end of the ¢rst 7-h session (Fig. 9A), but the non-embedded and embedded growth functions subsequently were very stable throughout the remainder of the 10-day regimen, and nearly the entire range of the stimulus pulse amplitude is represented in the embedded growth function. In this animal, the stimulus charge/phase reached 7.5 nC/phase, but the neurons and neuropil dorsal to the tip site are undamaged. We have shown previously that 7 h of continuous pulsing at 500 Hz, and at a charge per phase greater than 3 nC will induce microvaculations in the myelinated axons adjacent to the intranuclear microelectrode (McCreery

Fig. 9. A: The e¡ect of transposing the entire range of the stimulus pulse amplitudes, to compensate for an elevated electrical threshold (due to the microelectrode tip being slightly ventral to the cochlear nucleus. The threshold of the AER initially was slightly below 30 WA, and during the 10-day stimulation regimen, the arti¢cial voice modulated the 150-Ws pulses between 30 and 50 WA, as indicated on the abscissa. Considerable SIDNE developed during ¢rst 7-h session, but the growth function were stable throughout the remainder of the 10 days, and the embedded growth functions continued to span most of the range of stimulus amplitudes. B: An unsuccessful attempt to compensate for the stimulation-induced depression of the embedded growth function, by transposing the range of the stimulus pulse amplitudes. At the start of the ¢rst 7-h session, the arti¢cial voice signal modulated the stimulus pulses between 6 and 20 WA, which was a good match to the embedded RGF. By the end of the day of stimulation, the lower end of the growth function had become depressed. Our attempt to recover the original response range, by transposing the range of the stimulus upwards, only engendered further depression of the growth function.

HEARES 3558 4-10-00

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

233

et al., 1994). This did not occur in this animal, probably because the stimulus pulse rate was lower (250 Hz) and because the electrode tip was slightly outside of the nucleus and was not surrounded by myelinated axons. In contrast to the situation depicted in Fig. 9A, it will be di¤cult to compensate for SIDNE and SANR by transposing the range of the stimulus. In Fig. 9B, the threshold of the evoked response initially was low (approximately 6 WA). The arti¢cial voice signal modulated the 250-Hz stimulus between 6 and 20 WA (initially a good match to the embedded RGF), but by the end of the 7 h of stimulation, the embedded growth function had shifted downward and to the right, in the characteristic manner, and most of the response was to stimuli of 10 WA or greater. On the third day, the arti¢cial voice signal was adjusted to modulate the stimulus between 10 and 24 WA. By the end of the day, the embedded growth function had become shifted further, and the dynamic range of the embedded response was nearly identical to that prior to transposition of the range of stimulus pulse amplitudes. The experiment was repeated with four other electrodes, all with essentially the same result. 3.2. The e¡ect of stimulus pulse duration on the severity of SIDNE and SANR Intranuclear microstimulation probably will be easier to implement with existing implantable clinical stimulators by using stimulus pulses of relatively short duration and high amplitude. These stimulators were designed to pulse larger electrodes, and therefore to operate over a greater range of currents than 20 or 30 WA. To examine the e¡ect of prolonged stimulation with the short stimulus pulses, it was ¢rst necessary to establish the correspondences between the amount of neuronal activity evoked by the short and long pulses. Fig. 10A shows AERs recorded in the IC while stimulating in the contralateral PVCN with 150 Ws/phase or with 40 Ws/phase pulses. The ¢rst (P1^N1) and second (P2^N2) components are similar in duration and amplitude, when the stimulus pulse amplitudes are scaled appropriately. Fig. 10B was generated from the amplitude of the ¢rst component of the AERs evoked by a microelectrode in the PVCN. The abscissa is the amplitude of the 40-Ws stimulus pulses that evoked a response of the amplitude indicated by the values printed adjacent to the correspondence curve. The ordinate is the amplitude of the 150-Ws pulse that evoked a response of the same amplitude. The best (least squares) ¢t to the correspondence curve has a slope of 2.2, and this value was used as the neuronal excitability scaling factor. Six cats received stimulation regimens in which the pulse duration was 40 Ws. For 13 microelectrodes from

Fig. 10. A: AERs recorded in a cat's right IC and elicited by microstimulation in the left PVCN using charged-balanced, cathodic-¢rst current pulses that were either 150 Ws/phase (upper trace) or 40 Ws/ phase in duration (lower trace). B: The procedure used to compute the neuronal excitability scaling factor (`NESF') that was used to re-scale the stimulus amplitude range for the 40-Ws pulses. The numbers adjacent to the plots indicate the amplitude of the ¢rst component of the AER (P1^N1 in A) evoked from this microelectrode by 40- or 150-Ws current pulses having the amplitudes indicated on the abscissa and ordinate, respectively. The NESF is the slope (2.2) of the least mean squares ¢t to the data points.

these animals, the scaling factor ranged from 2.2 to 2.6 with a mean of 2.4. If the 40-Ws pulses were scaled so as to deliver the same charge per phase at the 150 Ws pulses, the factor would be 3.75, so on this basis the short-duration pulses are more e¤cient. For six microelectrodes, we used the mean value of 2.4 to scale the arti¢cial voice signal, so the range of 6^20 WA (which produced moderate SIDNE and SANR with the 150-Ws

HEARES 3558 4-10-00

234

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

Fig. 11. A: The iSIDNE for 13 microelectrodes during multi-day stimulation regimens in which the stimulus pulse duration was 40 Ws/phase. The index is shown at the end of the ¢rst day and last day of the 10-, 11-, or 12-day regimens. During the 7-h daily sessions, the arti¢cial voice signal modulated the pulse amplitudes between 14 and 48 WA (six microelectrodes) or between 14 and 63 WA (seven microelectrodes). Open and ¢lled symbols represent the data from the individual electrodes, and the group means, respectively. Error bars are the group standard deviations, at the beginning and end of the multi-day regimens. B: The iSANR for eight of the same microelectrodes depicted in A. C: iTOTAL for the eight microelectrodes.

pulses) was scaled to range from 14 to 48 WA. For the other seven microelectrodes, the arti¢cial voice modulated the pulse amplitudes between 14 and 63 WA (63 WA being the maximum output of our microstimulator). In all cases, the baseline stimulus current was 14 WA, which is close to the response threshold when the pulse duration is 40 Ws. The regimen of long-term stimulation otherwise was as described previously; the microelectrodes were each pulsed at 250 Hz in the interleaved mode, for 7 h/day, on 10 to 12 successive days, and

the stimulus pulse amplitude was modulated according to the arti¢cial voice signal, with a duty cycle of 50%. Fig. 11A shows the iSIDNE at the end of the ¢rst and last day of the regimen, for the 13 microelectrodes. The indices were computed as illustrated in Fig. 5B. Fig. 11B shows the iSANR for eight of these microelectrodes, and Fig. 11C shows the sum of the SIDNE and SANR for the same eight. When the arti¢cial voice signal modulated the signal between 14 and 48 WA, the indices were comparable to those when the stimulus

HEARES 3558 4-10-00

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

235

Fig. 12. A: A histologic section through the tip site (T) of a microelectrode that was pulsed for 7 h/day on 11 consecutive days. The stimulus pulse duration was 40 Ws/phase and the stimulus amplitude was modulated between 14 and 48 WA. The neurons (N) and neuropil close to the tip site appear normal. B: The site of the tip (T) of a microelectrode pulsed for 7 h/day on 11 consecutive days over the range of 14^63 WA. The neurons (N) and neuropil adjacent to the tip site appear normal and healthy, but the gliotic capsule surrounding the tip is slightly thicker and more cellular than those surrounding the tips of electrodes pulsed at a lower current density. Nissl stain. Bar = 50 Wm.

ranges from 6 to 24 WA and the pulse duration was 150 Ws. (Fig. 6). At the end of the ¢rst day, and at the end of the last day of the regimens, there was signi¢cantly more SIDNE when the arti¢cial voice signal modulated the stimulus between 14 and 63 WA, vs. 14 and 48 WA (P 6 0.01, t-test with unequal n). For both ranges of the stimulus, the SIDNE was not signi¢cantly di¡erent between the ¢rst day, and the last day of the regimen (P s 0.5 for both ranges). The SANR was comparable for both ranges of the arti¢cial voice signal (Fig. 11B). iTOTAL was slightly greater when the stimulus ranged from 14 to 63 WA (Fig. 11C). That di¡erence was not signi¢cant (P = 0.19 on the ¢rst day of the regimen, P = 0.14 on the last day) but this may have been due to the low power of the t-test, due to the small number of electrodes in each group. Fig. 12A shows the site of the tip of a pulsed electrode from cat CN131 in which the stimulus pulses were 40 Ws/phase in duration, and the pulse amplitude was

modulated between 14 and 48 WA by the arti¢cial voice signal. The electrode was pulsed for 7 h on each of 11 consecutive days, and the cat was killed 107 days after electrode implantation. The neuropil and neurons close to the tip appear to be normal. There were no stimulation-induced histologic changes near any of the tip sites for which the amplitude of the 40-Ws pulses reached 48 WA. When the 40-Ws pulses reached 63 WA, the histologic ¢ndings were a bit more ambiguous. Fig. 12B shows the tip site of a microelectrode that was pulsed on 11 consecutive days, and in which the arti¢cial voice modulated the stimulus between 14 and 63 WA. The nearby neurons and neuropil appear quite normal, but the gliotic sheath around the electrode tip is slightly thicker and more cellular than the sheath surrounding the site of the unpulsed tip (Fig. 12C). The tip's vestment of glial cells (probably astrocytes) are somewhat more darkly stained. We have observed these changes in the glial sheath in at least half

HEARES 3558 4-10-00

236

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

of the electrodes that were pulsed for 10^12 days with these parameters. It is notable that the stimulus charge per phase and charge density associated with the gliotic reaction were lower than those used when the pulse amplitude was 32 WA and the pulse duration 150 Ws/ phase (parameters which did not produce such a reaction) but the current density was higher. As noted above, the higher stimulus amplitudes did induce more SIDNE, but it did not become more severe over the course of the multi-day regimen, which suggest that any ongoing cellular reaction did not a¡ect neuronal excitability. However, prudence would dictate that when the short-duration, high-current pulses are used, the electrodes should have larger surface areas. Indeed, we have not observed this gliotic reaction in some very recent experiments in which the electrode surface area was 2000 Wm2 . This will be the topic of a future publication. 3.3. Summary of the stimulation protocols The iSIDNE, iSANR, and iTOTAL, are summarized in Table 1, for the four stimulation protocols that we evaluated. iSIDNE is given as the percent change relative to the pre-stimulus RGF. iSANR is given as the percent change relative to the non-embedded growth function acquired at the end of the daily session (see Fig. 5B). iTOTAL is the percent change relative to the prestimulus non-embedded growth function. Values are means þ standard deviation. The number of electrodes in the group is indicated in parentheses. The iSIDNE are signi¢cantly di¡erent for protocols A and B, and also for C and D. (P 6 0.001 for the contrast between A and B, P 6 0.01 for the contrast between C and D). For none of the protocols was the SIDNE signi¢cantly di¡erent between the end of the ¢rst day and last day of the regimen (P s 0.5). Between protocols A and B, iTOTAL also was signi¢cantly different, both on the ¢rst and last days (P 6 0.01). Between protocols C and D, iTOTAL was not signi¢cantly di¡erent (P = 0.19 on the ¢rst day, P = 0.14 on the last day) but this may have been due to the low power of the t-test, due to the small number of electrodes in each group.

4. Discussion The goals of this study were to help to establish the feasibility of extended microstimulation within the human VCN, and to begin to de¢ne the range of stimulus parameters that can be used safely for this application. Tissue injury is a potential hazard of prolonged intranuclear microstimulation at high charge per phase and pulsing rates, but even when histologically detectable tissue injury does not occur, the electrical excitability of the neurons near the microelectrodes may become suppressed. We have partitioned the stimulation-induced reduction of neuronal excitability into a component which persists for many hours or even days after the end of the high-rate stimulation (SIDNE) and SANR, which is apparent only during the high-rate stimulation (250 Hz). SIDNE is measured as a persisting shift in the RGFs generated by a low-rate probe stimulus. However, both SIDNE and SANR are strongly dependent upon the pulsing rate during the 7-h daily sessions (Fig. 5), and SIDNE, at least, requires several hours to develop fully (McCreery et al., 1997). A stimulation protocol should be designed so as to minimize the sum of the SIDNE and SANR, so that most of the neurons near the microelectrodes can continue to respond to most of the amplitude range of the stimulus pulses. In our cat model, and with a stimulation regimen spanning 10^21 days, none of the protocols induced histologically detectable neuronal injury, but all induced SIDNE and SANR. Neural injury was not observed in any of the animals, even when the maximum stimulus charge per phase exceeded 3 nC. This probably can be attributed to the relatively low stimulus frequency (250 Hz) and the fact that only a fraction of the stimulus pulses approached the maximum amplitude. Long-lasting SIDNE has been described for the cerebral cortex (McCreery et al., 1986), the cochlear nucleus (McCreery et al., 1997) and the auditory nerve (Tykocinski et al., 1995, 1997; Huang et al., 1998, Huang and Shepherd, 1999). These studies also noted the dependence of SIDNE upon stimulus pulse amplitude, pulsing rate and the stimulus duty cycle. In addition to the reduction of the dynamic range of the neuro-

Table 1 Protocol

Stimulus amplitude (range, WA)

Pulse duration (Ws)

iSIDNE, ¢rst day (%)

iSIDNE, last day (%)

iSANR, ¢rst day (%)

iSANR, last day (%)

iTOTAL, ¢rst day (%)

iTOTAL, last day (%)

A B C D

6^20 6^32 14^48 14^63

150 150 40 40

8.1% þ 10 (12) 65 þ 24 (4) 2.4 þ 1.7 (6) 19 þ 13 (8)

5.3 þ 11 (12) 73 þ 26 (4) 0.6 þ 2.1 (6) 20 þ 13 (8)

19 þ 7.5 (4) 32 þ 18 (4) 16 þ 9.7 (4) 21 þ 8.1 (4)

19 þ 8.1 (4) 43 þ 21 (4) 12 þ 12 (4) 18 þ 6.1 (4)

22 þ 5.6 (4) 76 þ 16 (4) 25 þ 3.3 (4) 30 þ 4.4 (4)

23 þ 10 (4) 80 þ 13 (4) 25 þ 8.4 (4) 33 þ 4.8 (4)

HEARES 3558 4-10-00

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

nal response and its implications for the performance of a clinical prosthesis, the relationship between SIDNE and incipient neural injury is yet to be fully elucidated, and for both of these reasons, it is prudent to design the stimulation protocol so as to minimize this phenomenon. Similarly, it is unclear what mechanisms might underlie the SANR, which is detected as a depression of the growth functions of the evoked potentials that are recorded in the IC during the stimulation at the full rate or 250 Hz. SANR may be a manifestation of forward masking, although we have not documented the timecourse of SANR su¤ciently to precisely establish the correspondence between the two phenomena. Therefore, we prefer the more general term `short-acting neuronal refractivity'. Goldstein and Kiang (1958) noted that a compound action potential evoked by a transient stimuli could be considered to be composed of the waveform of the individual action potentials convolved with the post-stimulus time histogram of the neuronal population contributing to the compound response. Thus, a portion of the SANR might be due to a reduction in the amplitude of the individual action potentials or to increased temporal dispersion of the individual action potentials. However, from the perspective of this study's objectives, it is most instructive to emphasize the portion of the neuronal response that remains after SIDNE and the superimposed SANR have been subtracted, rather than to speculate about the processes that underlie these phenomena. For two of the stimulation protocols that we investigated (150 Ws/phase biphasic pulses modulated between 6 and 20 WA, and 40 Ws/phase pulses modulated between 14 and 48 WA) the combined e¡ect of SIDNE and SANR was to depress the RGFs by an average of 25% (Table 1) and by a maximum of no more than 35% (Figs. 6C and 11C). Also, with these stimulus parameters, the SIDNE and SANR in most cases did not change signi¢cantly between the end of the ¢rst and last day of the multi-day regimens. The combined e¡ect of SIDNE and the superimposed SANR is to depress the neuronal response near threshold, and thereby, to depress the population response over the entire amplitude range of the stimulus pulses (Fig. 5B). However, the SIDNE and SANR may cause the greatest degradation of the performance of a clinical device at the low end of the amplitude range, where a partial `dead band' may develop. SIDNE apparently is related to the neuronal activity induced by the stimulation (McCreery et al., 1997) and when the stimulus pulses are distributed over a range of amplitudes, as in Fig. 2, the neurons with the lowest threshold (presumably those closest to the electrode) will be excited by nearly every current pulse, while those with the higher thresholds will be excited less frequently and thus

237

sustain less depression. The consequence of the dead band may be ameliorated by the logarithmic amplitude compression of the auditory signal, which causes most of the stimulus pulse amplitudes to be near the upper end of the range (Fig. 2), and by the inclusion in the sound processor of an automatic gain control, which will allow optimal utilization of the remaining dynamic range of neuronal responses. If the dead band does develop in a clinical device, and if the patients' performance su¡ers, then it may be possible to transpose the entire range of stimulus amplitudes upwards by a few WA, and at least partly out of the dead band. However, Fig. 9B indicates that this strategy may be of limited value, since increasing the stimulus amplitude will exacerbate the SIDNE. Another strategy would be to reduce the pulsing rate (Fig. 5, and McCreery et al., 1997), but this may adversely a¡ect the patient's performance. However, if the responses of the neurons in the feline and human ventral cochlear nuclei are similar, then the depression of neuronal excitability during prolonged stimulation may not seriously impair the patient's ability to comprehend speech. Fig. 9A illustrates that it is possible to compensate for an initially high neuronal response threshold (in this case, due to the stimulating microelectrode being slightly out of position), by shifting the entire range of stimulus amplitude, so that acoustic silence corresponds to the neuronal response threshold (or, with a clinical device, the patient's perceptual threshold). In a clinical prosthesis, this strategy should be reserved for instances in which the threshold of the electrically evoked auditory percepts are elevated, since prolonged stimulation at a high charge per phase carries a risk of injury to myelinated axons that pass within approximately 100 Wm of the electrode tips (McCreery et al., 1994). In a human patient, an electrode tip that penetrates completely through the cochlear nucleus would reside in the inferior cerebellar peduncle, and the consequences of possibly damaging the axons that pass very close to the electrode are uncertain. Certainly, these animal studies leave unanswered a number of questions regarding the expected level of performance by a patient with a clinical prosthesis that is based on intranuclear microstimulation. It is not certain that the neurons in the PVCN of a deafened human will respond to prolonged electrical stimulation in the same manner as those of a cat with normal hearing. The real-world equivalent of stimulus duty cycle (the portion of time in which high-level sound is present) is nearly impossible to estimate, and a patient who lives or works in a noisy environment may experience more SIDNE. The e¡ects of very long-term microstimulation in the VCN have yet to be determined. Also unknown are the types of auditory percepts that would be elicited by stimulation at only a few loci along

HEARES 3558 4-10-00

238

D.B. McCreery et al. / Hearing Research 149 (2000) 223^238

the tonotopic gradient of the VCN, and how the recruitment of this highly localized neuronal population will covey loudness. However, our studies do provide some support for clinical trials of a device that would include the surface electrodes used in the present ABI, and also an array of intranuclear penetrating microelectrodes. Acknowledgements We extend special thanks to J. Philip Mobley, M.S.E.E., formerly of the House Ear Institute, for his patient assistance with the sound pressure measurements, and for providing us with a digital tape and documentation of the CCITT arti¢cial voice signal. This work was supported by Contracts NO1-DC-52105 and NO1-DC-8-2102 from the National Institutes of Health (NIDCD). The animal studies were conducted using a protocol approved by the HMRI Institutional Animal Care and Use Committee, and the study was conducted according to the committee's standards. References The arti¢cial voice signal is described in Report P50 of Working Party Q.12/XII, The International Telegraphic and Telephony Consultive Convention (CCITT), January 1988, Geneva. Brackmann, D.E., Hitselberger, W.E., Nelson, R.A., Moore, J.K., Waring, M., Portillo, F., Shannon, R.V., Elischi, F., 1993. Auditory brainstem implants. I: issues in surgical implantation otolaryngology. Head Neck Surg. 108, 624^634. Brill, S.M., Gostottner, W., Helms, J., von Ilberg, C., Baumgartner, W., Muller, J., Kiefer, J., 1997. Optimization of channel number and stimulation rate for the fast continuous interleaved sampling strategy in the COMBI 40+. Am. J. Otol. 6 (Suppl.), S104^S106. Brunso-Bechtold, J.K., Thompson, G.C., Masterton, R.B., 1981. HRP study of the organization of the auditory a¡erents ascending to central nucleus of the inferior colliculus in cat. J. Comp. Neurol. 197, 705^722. Cant, N.B., 1982. Identi¢cation of cell types in the anteroventral cochlear nucleus that project to the inferior colliculus. Neurosci. Lett. 32, 141^246. Eisenberg, L.S., Maltan, A.A., Portillo, F., Mobley, J.P., House, W.F., 1987. Electrical stimulation of the auditory brain stem in deafened adults. J. Rehabil. Res. Dev. 24, 9^22. Fishman, K., Shannon, R.V., Slattery, W.H., 1997. Speech recognition as a function of the number of electrodes used in the SPEAK cochlear implant speech processor. J. Speech Hear. Res. 40, 1201^ 1215. Goldstein, M.H., Kiang, N.Y., 1958. Synchrony of neural activity in the electric response evoked by transient acoustic stimuli. J. Acoust. Soc. Am. 30, 107^114. Huang, C.Q., Shepherd, R.K., Seligman, P.M., Clark, G.M., 1998. Reduction in excitability of the auditory nerve following acute electrical stimulation at high stimulus rates: III: Capacitive versus

non-capacitive coupling of the stimulating electrodes. Hear. Res. 116, 55^64. Huang, C.Q., Shepherd, R.K., 1999. Reduction in excitability of the auditory nerve following acute electrical stimulation at high stimulus rates: IV: e¡ects of stimulus intensity. Hear. Res. 132, 60^68. McCreery, D.B., Bullara, L.A., Agnew, W.F., 1986. Neuronal activity evoked by chronically implanted intracortical microelectrodes. Exp. Neurol. 92, 147^161. McCreery, D.B., Yuen, T.G.H., Agnew, W.F., Bullara, L.A., 1992. Microstimulation in the cochlear nucleus of the cat with chronically implanted microelectrodes: histologic and physiologic e¡ects. Hear. Res. 62, 42^56. McCreery, D.B., Yuen, T.G.H., Agnew, W.F., Bullara, L.A., 1994. Stimulation parameters a¡ecting tissue injury during microstimulation in the cochlear nucleus of the cat. Hear. Res. 77, 105^115. McCreery, D.B., Yuen, T.G.H., Agnew, W.F., Bullara, L.A., 1997. A characterization of the e¡ects on neuronal excitability resulting from prolonged microstimulation with chronically implanted microelectrodes. IEEE Trans. Biomed. Eng. 44, 931^939. McCreery, D.B., Shannon, R.V., Moore, J.K., Chategee, M., 1998. Accessing the tonotopic organization of the ventral cochlear nucleus by intranuclear microstimulation. IEEE Trans. Rehabil. Eng. 23, 391^399. Moore, J.K., Osen, K.K., 1979. The cochlear nuclei in man. AAER J. Anat. 154, 393^417. Moore, J.K., 1987. The human auditory brain stem: a comparative view. Hear. Res. 29, 1^32. Otto, S.R., Shannon, R.V., Brackmann, D.E., Hitselberger, W.R., Staller, S., Menapace, C., 1997. The multichannel auditory brainstem implant performance in 20 patients. Otolaryngol. Head Neck Surg. 118, 291^306. Parkin, J.L., Eddington, D.K., Orth, J.L., Brackmann, D.E., 1985. Speech recognition with multichannel cochlear implants. Otolarynolgol. Head Neck Surg. 93, 639^645. Robblee, L.S., Lefko, J., Brummer, S.B., 1983. Activated Ir: an electrode suitable for reversible charge injection in saline solution. J. Electrochem. Soc. 130, 731^733. Shannon, R.V., 1989. Threshold functions for electrical stimulation of the human cochlear nucleus. Hear. Res. 40, 173^178. Shannon, R.V., Otto, S.R., 1990. Psychophysical measures from electrical stimulation of the human cochlear nucleus. Hear. Res. 47, 159^168. Shannon, R.V., Fayad, J., Moore, J.K., Lo, W., O'Leary, M., Otto, S.R., Nelson, R.A., 1993. Auditory brainstem implant: II Postsurgical issues and performance. Otolaryngol., Head Neck Surgery 108, 635^643. Shannon, R.V., Zeng, F.-G., Wygonski, J., Kamath, V., Ekelid, M., 1995. Speech recognition with primarily temporal cues. Science 270, 303^304. Tykocinski, M., Shepherd, R.K., Clark, G.M., 1995. Reduction in excitability of the auditory nerve following electrical stimulation at high stimulus rates. Hear. Res. 88, 124^142. Tykocinski, M., Shepherd, R.K., Clark, G.M., 1997. Reduction in excitability of the auditory nerve following electrical stimulation at high stimulus rates: II: Comparison of ¢xed amplitude with amplitude modulated stimuli. Hear. Res. 112, 147^157. Warr, W.B., 1972. Fiber degeneration following lesions in the multipolar and globular cell areas in the ventral cochlear nucleus of the cat. Brain Res. 40, 247^270. Wilson, B.S., Lawson, D., Zerbi, M., Finley, C.C., Wolford, R.D., 1995. New processing strategies in cochlear implantation. Am. J. Otol. 16, 669^675.

HEARES 3558 4-10-00