Characterization of Temporal Interactions in the Auditory Nerve of Adult and Pediatric Cochlear Implant Users

Transcription

1 University of Iowa Iowa Research Online Theses and Dissertations Summer 2013 Characterization of Temporal Interactions in the Auditory Nerve of Adult and Pediatric Cochlear Implant Users Aayesha Narayan Dhuldhoya University of Iowa Copyright 2013 Aayesha Narayan Dhuldhoya This dissertation is available at Iowa Research Online: Recommended Citation Dhuldhoya, Aayesha Narayan. "Characterization of Temporal Interactions in the Auditory Nerve of Adult and Pediatric Cochlear Implant Users." PhD (Doctor of Philosophy) thesis, University of Iowa, Follow this and additional works at: Part of the Speech and Hearing Science Commons

2 CHARACTERIZATION OF TEMPORAL INTERACTIONS IN THE AUDITORY NERVE OF ADULT AND PEDIATRIC COCHLEAR IMPLANT USERS by Aayesha Narayan Dhuldhoya A thesis submitted in partial fulfillment of the requirements for the Doctor of Philosophy degree in Speech and Hearing Science in the Graduate College of The University of Iowa August 2013 Thesis Supervisor: Professor Paul J. Abbas

3 Graduate College The University of Iowa Iowa City, Iowa CERTIFICATE OF APPROVAL PH.D. THESIS This is to certify that the Ph.D. thesis of Aayesha Narayan Dhuldhoya has been approved by the Examining Committee for the thesis requirement for the Doctor of Philosophy degree in Speech and Hearing Science at the August 2013 graduation. Thesis Committee: Paul J. Abbas, Thesis Supervisor Carolyn J. Brown Shawn S. Goodman Marlan Hansen Christopher W. Turner

4 TABLE OF CONTENTS LIST OF TABLES... v LIST OF FIGURES... vi LIST OF ABBREVIATIONS...viii CHAPTER 1 INTRODUCTION Temporal structure of speech: Coding and relevance to perception Temporal coding in cochlear implants Relevance to speech coding Psychophysical and speech perception measures Adaptation and Recovery Proposed study Rationale, Proposed study and Criterion measure Specific Research Questions and Hypothesis CHAPTER 2 REVIEW OF LITERATURE Two-pulse Paradigms Integration Refractoriness Integration Refractoriness and Alternation High-Rate stimulation Stochastic Responses Increased Dynamic Range Adaptation Amplitude Modulation Recovery / Pulse train forward masking CHAPTER 3 METHODS Study Participants General Procedures Stimulation and Recording Set-Up Behavioral and ECAP measures Bases of Experimental Paradigm The electrically evoked compound action potential (ECAP) Issues in ECAP measurement: Noise and stimulus artifact Basic Subtraction technique Modified Subtraction technique Stimulus and Recording parameters ii

5 3.4 Specific Testing Procedures Behavioral T and C levels ECAP thresholds and amplitude growth function Baseline ECAP and artifact template measurements Two pulse temporal interactions: Refractory Recovery Multi pulse temporal interactions: Adaptation Multi pulse temporal interactions: Recovery from Adaptation Post-hoc analyses of stimulus artifact CHAPTER 4 RESULTS Behavioral and ECAP Baseline Measures Two pulse temporal interactions (Refractory Recovery) Analyses of categorical data Analyses of continuous data Discussion: Two-pulse interactions Multi-pulse temporal interactions: Adaptation Normalization and Analyses Onset Response (Long-term Adaptation) Steady-state Responses Discussion: Adaptation Multi-pulse temporal interactions: Pulse train forward masking / Recovery from Adaptation Long-term adaptation (Normalized probe alone measures) Recovery from Adaptation (Pulse train forward masking) Components of Recovery Level and Age Effects Relationship between single- and multi-pulse forward masking Discussion: Pulse train forward masking CHAPTER 5 DISCUSSION AND CONCLUSIOMS APPENDIX A. ARTIFACT MEASUREMENT AND REMOVAL APPENDIX B. DEMOGRAPHIC DATA, BEHAVIORAL AND ECAP BASELINE MEASURES APPENDIX C. AMPLITUDE NORMALIZATION PROCEDURE APPENDIX D. NORMALIZED MEAN ECAP AMPLITUDE AT VARIOUS MASKER-PROBE-MPI COMBINATIONS IN CHILDREN AND ADULTS: TWO PULSE DATA APPENDIX E. NORMALIZED MEAN ECAP AMPLITUDE AT VARIOUS STIMULUS LEVELS IN CHILDREN AND ADULTS: PULSE TRAIN DATA iii

6 APPENDIX F. MODEL OF POPULATION RESPONSE REFERENCES iv

7 LIST OF TABLES Table B.1 Hearing loss and implant information in children Table B.2 Hearing loss and implant information in adults Table B.3 Behavioral and ECAP baseline measures in children Table B.4 Behavioral and ECAP baseline measures in adults Table D.1 Table E.1 Normalized mean ECAP amplitudes at various masker-probe-mpi combinations in children and adults: Two pulse data Normalized mean ECAP amplitudes at various stimulus levels in children and adults: Pulse train data v

8 LIST OF FIGURES Figure 1.1 Neural coding of speech... 3 Figure 3.1 Stimulation and recording set up for ECAP measures Figure 3.2 Basic Subtraction technique: Stimulus Paradigm Figure 3.3 Basic Subtraction technique: Recording Figure 3.4 Modified Subtraction technique Figure 4.1 Stimulus levels in behavioral and ECAP measures of individual subjects Figure 4.2 ECAP amplitudes at various stimulus levels in individual subjects Figure 4.3 Figure 4.4 Figure 4.5 Normalized ECAP amplitudes at various stimulus levels in individual subjects Mean normalized ECAP amplitudes at various masker-probe-mpi (twopulse) combinations in the two age groups Individual normalized ECAP amplitudes at various ratios of the baseline probe to masker ECAP amplitude Figure 4.6 LOESS curve fits Figure 4.7 Exemplar ECAPs in response to pulse trains Figure 4.8 Figure 4.9 Average and individual normalized ECAP amplitudes at onset and steady state for pulse train stimuli Individual ECAP amplitudes for the probe pule in the two- and multipulse experiments Figure 4.10 Average normalized ECAP amplitudes for the probe alone in the twoand multi-pulse experiments Figure 4.11 Individual ECAP amplitudes at asymptote Figure 4.12 Relationship between steady-state adaptation (re: onset) and the longterm adaptation Figure 4.13 Relationship between ECAPs in two-pulse and multi-pulse stimulus paradigms vi

9 Figure 4.14 Recovery in children Figure 4.15 Recovery in adults Figure 4.16 Patterns of recovery described by possible underlying processes Figure 4.17 Estimated time constants of various processes Figure 4.18 Level effects on pulse train forward masking Figure 4.19 Recovery functions categorized by baseline ECAP amplitude Figure 4.20 Relationship between pulse train and single pulse forward masking Figure 4.21 Adaptation, conditioning and the relationship between them Figure C.1 Amplitude normalization Figure F.1 Estimated unit potential: Waveforms and spectra Figure F.2 Figure F.3 Figure F.4 Figure F.5 Simple model of convolution with hypothetical unit potentials and firing distributions Measured and reconvolved ECAP waveforms from the two-pulse study with computed firing distributions Measured and reconvolved ECAP waveforms from the adaptation study with computed firing distributions Measured and reconvolved ECAP waveforms and computed firing distributions during recovery from adaptation vii

10 LIST OF ABBREVIATIONS ANF CAP CI CL C-level db DPT ECAP Hz IPI MPI ms msec Auditory nerve fiber Compound action potential Cochlear implant Current level Maximum Comfort(able) level Decibels Desynchronizing pulse train Electrically evoked compound action potential Hertz Inter-pulse interval Masker-probe interval Millisecond Millisecond µs Microsecond µv Microvolts pps PST T-level Pulses per second Post-stimulus time Threshold level viii

11 1 CHAPTER 1 INTRODUCTION The envelope cues within the speech signal, particularly the onsets of speech segments, have a robust neural representation that is preserved and, possibly, enhanced along the auditory pathways. Cochlear implants encode the temporal envelope of various spectral bands of the speech signal in the amplitude of the individual pulses that are delivered to different electrodes in the array. High-rate pulse trains allow greater temporal detail of the speech signal to be delivered to the nerve; however, they may induce greater adaptation and possibly, incomplete recovery of neural responsiveness between speech segments and thus, have the potential to obfuscate the neural code for speech onsets. This study proposes to characterize neural adaptation and recovery thereof in users of current cochlear implant systems in order to improve our understanding of the neural coding of speech in this population particularly with regard to the developmental effects, if any, of these phenomena Temporal structure of speech: Coding and relevance to perception The temporal structure of speech provides important linguistic contrasts and may be described by a three-way classification system (Rosen, 1992). The slow temporal envelope cues (2 to 50 Hz) convey consonantal manner as well as some information about voicing and the rhythm, syllabicity and stress patterns of speech. Periodicity cues (50 to 500 Hz) signal the occurrence and frequency of vocal fold vibration. Besides voicing, they convey the manner of articulation, prosody, i.e., intonation and stress, and non-linguistic information, viz., speaker identity. Fine structure fluctuations at 600 to 10,000 Hz relate to the spectral shape; they cue place of articulation and vowel identity.

12 2 The frequency analysis by the cochlea enables the transformation of the periodicity and fine structure cues into a (rate-)place code (Rosen, 1992), by the primary auditory neurons, that can be maintained by the tonotopic organization of the central auditory pathways. The phase locking ability of the primary auditory neurons allows for temporal coding of the periodicity and fine structure which may be useful in disambiguating spectral shape at stimulus levels for which the place code is degraded (Sachs & Young, 1979). Perception studies indicate that normal listeners use the rapid temporal information only in limited instances, e.g., complete (van Tasell, Soli, Kirby & Widin, 1987) or severe (Smith, Delgutte & Oxenham, 2002) spectral degradation of the speech signal, speech with limited contextual cues or presented in noisy environments (Zeng, Nie, Stickney, Kong, Vongphoe, Bhargave, Wei & Cao, 2005) and recognition of linguistic contrasts in tonal languages (Zeng et al, 2005) etc. Slow envelope cues, however, must be encoded temporally. Envelope features include intensity, duration, silent intervals, onsets and offsets of the signal (Delgutte, 1997; Rosen, 1992). They, too, may be particularly useful if spectral resolution is limited. Normal listeners achieve high speech recognition with just three spectrally contiguous bands of noise being modulated by the speech envelope (Shannon, Zeng, Kamath, Wygonzki & Ekelid, 1995) and speech intelligibility is similarly well preserved even with limited spectral detail if all the amplitude fluctuations below 8-10 Hz are maintained (Drullman, Festen & Plomp, 1994). Additionally, the intelligibility of noise corrupted speech has been shown to improve when syllable and word onsets/offsets are made accessible by reducing the low frequency energy in the time segments corresponding to obstruent consonants (Li & Loizou, 2008).

13 3 Figure 1.1: Neural code of speech. From Neural coding of the temporal envelope of speech: Relation to modulation transfer functions, by B. Delgutte, B.M. Hammond and P.A. Cariani, 1998, in Psychophysical and Physiological Advances in Hearing by A.R. Palmer, A. Reese, A.Q. Summerfield, and R Meddis (Eds), London: Whurr, p

14 4 Envelope cues, especially rapid spectral and amplitude changes or acoustic onsets, are prominently represented in the response of auditory nerve fibers or ANFs (Delgutte & Kiang, 1984). ANFs respond to rapid increases in energy at or near their characteristic frequency (CF) with a rapid increase in their firing rate followed by a decay in responsiveness due to adaptation (Smith, 1977) so that the post stimulus time (PST) histograms of ANFs show adaptation peaks at speech onsets (Delgutte, 1997). Indeed, onsets of speech segments gain prominence due to the large phasic and smaller tonic responses of the auditory nerve and the cochlear nucleus and due to the largely phasic responses of neurons in the inferior colliculus (Delgutte, Hammond & Cariani, 1998) as seen in Figure 1.1. The selective preservation of onsets may attest to their relevance in speech perception. Certainly, phasic responses to onsets at the cochlear nucleus and beyond may be the result of spatio-temporal integration of tonic input from the periphery and may not require a phasic response in the nerve. However, in order for an adequate tonic and, possibly, robust phasic nerve output to occur at onsets in continuous speech, rapid recovery of ANFs from adaptation by preceding speech segments may be essential. There may be other indirect evidence of the importance of onsets to speech perception. For instance, recovery from adaptation may be critical to tasks of gap detection in that a robust neural response may be required at the onset of the post gap marker. Furthermore, there appears to be an association between gap detection and speech recognition ability, i.e., older adults, as a group, have poorer gap detection thresholds and speech recognition in babble than do younger adults with similar hearing sensitivity (Snell & Frisina, 2000). If temporal precision in coding onsets is an important feature of the auditory system, then ANFs are likely to respond in unison at the onset of

15 5 phonetic segments. Hence, it is not surprising that the neural responses to clicks and rapid speech onsets may be remotely recorded as gross evoked potentials and that impaired speech recognition is associated with abnormal evoked potentials. For instance, abnormal acoustically evoked auditory brainstem responses (ABRs) and poor speech recognition are features of auditory dys-synchrony while electrically induced synchrony has been shown to be a good prognosticator for speech-language gains in children with cochlear implants (Teagle, Roush, Woodard, Hatch, Zdanski, Buss & Buchman, 2010). The co-existence of abnormal tone adaptation, ABRs and word recognition in neural pathologies may attest to the relevance of coding onsets. Finally, while the role of auditory deficits in learning disability (LD) is not clear, speech evoked ABRs in children with more severe LD show impaired representation of onsets (Banai, Nicol, Zerker & Kraus, 2005) Temporal coding in cochlear implants Most cochlear implants encode speech as trains of biphasic electric pulses that are amplitude modulated by the temporal envelope of the speech signal. Specifically, the signal is passed through a set of band pass filters with low cut-off frequencies between 200 to 400 Hz, and the temporal fluctuations at the output of various filters are used to modulate the amplitude of electric pulses delivered to different electrodes. It is hoped that the ANFs will, as a group, encode the modulation waveform Relevance to speech encoding Newer implant technology has the capability to deliver stimulation at rates up to 5,600 pulses per second (pps) per channel, so that all of the periodicity information and some fine structures cues may be encoded in speech processors (Wilson, Sun, Schatzer &

16 6 Wolford, 2004). Given their utility in maintaining the intelligibility of spectrally degraded speech, the fast temporal fluctuations may be beneficial to users of cochlear implants for whom the place information is limited. Thus, the relevant factor in providing an accurate temporal representation of the sound signal is no longer the capability of the implant but rather the temporal processing ability of the auditory nerve. It is not surprising that the response of ANFs to individual pulses is dependent not simply upon the level (i.e., charge) of the pulse but also upon the past history of stimulation and response. In acoustically sensitive ears, spontaneous neural discharges maintain ANFs in varying states of refractoriness, introducing variability in the instantaneous threshold values which in turn allows nerve fibers to respond to a stimulus with stochastic independence (Wilson, Finley, Zerbi & Lawson, 1994). In the absence of spontaneous activity, electrical stimulation is able to produce high driven rates over a very narrow dynamic range and neural spikes that are highly synchronized across fibers (Kiang & Moxon, 1972). Such synchrony may be reduced by the use of high rate pulse trains that are themselves modulated by the temporal envelope of the acoustic input (Wilson et al, 1994) or that serve to produce pseudo-spontaneous neural discharges which can then be modulated by a superimposed pulse train bearing the temporal envelope of the speech signal (Rubinstein, Wilson, Finley & Abbas, 1999). The latter strategy is said to use a desynchronizing pulse train (DPT). In either case, the reduction in synchrony of neural firing may allow for an increase in the dynamic range and a better neural representation of the temporal fluctuations in the speech signal. Pulse rates above 2000 pps are associated with smaller electrically evoked compound action potentials (ECAPs) indicating a reduction in synchronous driven

17 7 activity to individual pulses (Wilson, Finley, Lawson & Zerbi, 1997a; Wilson, Finley, Zerbi, Lawson & van den Honert, 1997b). They also increase the dynamic range by lowering fiber thresholds due to current integration at the neural membrane (Zhang, Miller, Robinson, Abbas & Hu, 2007) and by allowing very high discharge rates to be attained with increasing stimulus levels at least in acutely deafened animals (Javel & Shepherd, 2000; Shepherd & Javel, 1997). Spike activity induced by DPT at 5000 pps has the characteristics of spontaneous like discharges (Litvak, Smith, Delgutte & Eddington, 2003a). Also, the addition of a DPT is seen to reduce the slope of an ECAP growth function for a sinusoid (Runge-Samuelson, Abbas, Rubinstein, Miller & Robinson, 2004) and to improve the representation of modulations in ANF period histograms (Litvak et al, 2003a, Litvak, Delgutte & Eddington, 2003b) and of the fundamental frequency of a vowel in the human ECAP (Wilson, Lawson, Muller, Tyler & Kiefer, 2003) Psychophysical and speech perception measures While the psychophysical dynamic range has been found to increase with pulse rate (Galvin & Fu, 2005, 2009; Kreft, Donaldson & Nelson, 2004; Pfingst, Xu & Thompson, 2007) and the DPT has been shown to increase the dynamic range as well as produce a more gradual increase in the loudness of a sinusoid (Hong & Rubinstein, 2003), concomitant increases in the number of discriminable intensity steps have not been evident (Kreft et al, 2004; Galvin & Fu, 2009). Also, modulation detection, known to be predictive of speech recognition ability (Cazals, Pelizzone, Saudan & Boex, 1994; Fu, 2002), was found to worsen for slow modulations (Galvin & Fu, 2005, 2009; Pfingst et al, 2007) or failed to improve (Green, Faulkner & Rosen, 2012) at high carrier rates.

18 8 Importantly, speech perception studies yield conflicting reports about the advantage of high rate stimulation. Better speech recognition was reported using rates of 2,100 pps as compared to 800 pps or less (Loizou, Poroy & Dorman, 2000), for a rate of 1500 pps as compared to 400 and 800 pps (Verschuur, 2005) and for increasing the rate from 1000 to 4000 pps when listening with four spectral bands in noise (Nie, Barco & Zeng, 2006). In contrast, several other studies (Friesen, Shannon & Cruz, 2005; Fu & Shannon, 2000; Plant, Holden, Skinner et al, 2007; Shannon, Cruz & Galvin, 2011; Vandali, Whitford, Plant & Clark, 2000; Weber, Lai, Dillier, von Wallenberg & Killian, 2007) found no significant advantage to the use of high pulse rates. Limitations of central processing may explain this lack of benefit but temporal interactions, viz., adaptation and recovery, in the electrically stimulated nerve may also be relevant factors at high stimulation rates Adaptation and recovery Adaptation is a decrease in neural excitability and response probability over time. As observed with acoustic stimulation, physiologic adaptation is both per(i)stimulatory and post-stimulatory. The firing rate is maximal at the onset of sustained stimulation and adapts to a steady state level thereafter; spontaneous activity and responsiveness to subsequent stimulation are reduced for a brief period following the cessation of such stimulation resulting in forward masking effects (Smith, 1977). As suggested by Delgutte (1997), adaptation may have the important role of producing peaks in discharge rate that serve to enhance rapid spectro-temporal changes in speech. The formation of adaptation peaks in continuous speech presumably requires an initial robust response to syllable or word onsets, rapid adaptation as well rapid recovery from such adaptation.

19 9 Neural adaptation and forward masking from acoustic stimulation may differ from that due to electric stimulation. In acoustic hearing, synaptic adaptation, i.e., neurotransmitter depletion and/or desensitization of the post-synaptic receptor, reduces the input to the post-synaptic ANF and allows for membrane recovery. In contrast, persistent voltage changes at the neural membrane that result from high rate electrical stimulation as well as chemical changes due to higher driven rates may preclude the restoration of neural excitability until well after the withdrawal of stimulation. Constant amplitude electric pulses induce adaptation that has an exponential time course so that spike rate decreases rapidly within the first few milliseconds (ms) and reaches asymptote within a few hundred milliseconds (ms) after stimulus onset (Litvak, Delgutte & Eddington, 2001; Zhang et al, 2007). The degree and rate of adaptation increase with the rate of stimulation in animals (Abbas, Miller, Rubinstein, Robinson & Hu, 2001; Haenggeli, Zhang, Vischer, Pelizzone & Rouiller, 1998; Litvak et al, 2001; Zhang et al, 2007) and human implant recipients (Schmidt-Clay & Brown, 2007). Adaptation is reduced by slow sinusoidal modulations of the carrier pulse train but its time course is similar to that for an unmodulated pulse train (Hu, Miller, Abbas, Robinson & Woo, 2010). Recovery from adaptation by electrical stimulation also follows an exponential time course with an initial rapid and later slow phase in single fibers. Slower recovery, i.e., greater forward masking, was observed as the rate of pulsatile stimulation increased even without any or increased masker evoked activity (Miller, Woo, Abbas, Hu & Robinson, 2011). Although there are no reports of pulse train forward masking in human subjects, the animal data have an important implication. The use of high pulsatile rates

20 10 that are associated with greater forward masking could potentially reduce the salience of the onset response to successive speech segments. Consequently, high stimulus rates may, potentially, have deleterious effects on neural encoding and speech perception despite their ability to carry greater information about the acoustic input. Of relevance here is the observation of Vandali (2001) that cochlear implant users lack access to the information-rich albeit brief, low intensity cues such as burst envelopes and formant transitions, at least in part, due to the temporal and spatial masking of such cues. The duration of deafness may be a determinant of the amount of adaptation. There was a much higher incidence of total spike adaptation, particularly at high pulse rates, in chronically deaf as compared to acutely deaf animals (Miller C.A, Unpublished data) and for the same rate of stimulation, much lower spike rates could be achieved in short- and long-term deaf ears relative to acutely deafened ears (Shepherd & Javel, 1997). Indeed, the spatial and temporal interactions in the auditory nerve of animals that have intact hair cells or are acutely deafened prior to acoustic and/or electric stimulation may be very different from those observed in the congenital or long deafened ears of cochlear implant recipients. First, although, the diagnosis of implant recipients may be a cochlear lesion, a primary or secondary involvement of the auditory nerve is not improbable. Specifically, demyelination, loss of peripheral processes, shrinkage and loss of spiral ganglion cells that occur in the auditory nerve as a consequence of hair cell damage are particularly relevant to neural excitability (Shepherd & Javel, 1997) in response to both, initial and sustained stimulation. Secondly, the typical deafening procedure used in animal studies involves aminoglycoside ototoxicity which is known to be associated with relatively better preservation of the spiral ganglion (Nadol, 1997) and

21 11 hence, not likely to represent the widespread degeneration that could potentially occur in humans subjects. These factors may produce greater neural adaptation and forward masking in the implanted human ear as compared to animal models. It is also possible that some of the variability in spatio-temporal interactions may be related to demographics such as age. The immature auditory nerve of children may have different temporal response properties from that of the adult implant recipient. While the inner ear, cochlear nerve and the brainstem auditory pathways are developed by the second trimester of gestation, myelination and synapse strengthening continue during the first two post natal years as is evident in the maturational changes in the auditory brainstem response (ABR). An important feature of ABR maturation is the increased resistance of its faster components, particularly wave I amplitude, to rate effects (Jiang, Brosi & Wilkinson, 1998). While it is possible that the temporal response properties of the auditory nerve in children with cochlear implants undergo similar maturation as their hearing counterparts and no specific role has been ascribed to early stimulation, it is also possible that these properties in implanted children may remain quite different from those of post-lingually deafened adults especially if sound deprivation occurred during the critical years of neuromaturation. 1.3 Proposed Study Rationale and criterion measure In summary, cochlear implants use high rate pulse trains for better temporal sampling of the speech signal as well as to reduce the firing synchrony and increase the dynamic range within and across fibers so that the rapid amplitude modulations of speech may be encoded by the auditory nerve. However, high rate stimulation has not

22 12 consistently produced the expected improvements in intensity discrimination, modulation detection and speech perception. The benefit of high rate stimulation is predicated on its ability to desynchronize action potential firing without excessive reductions in the probability of firing across the neural population. However, greater adaptation and slower recovery of the neural response may accompany higher rates of pulsatile stimulation particularly in immature and/or disordered systems. The inability of the nerve to recover rapidly from prior stimulation and responses may be, potentially, detrimental to the encoding of speech envelope features, e.g., syllable and word onsets. The per- and post-stimulatory effects of adaptation produced by high rate pulse trains have not yet been characterized in cochlear implant users. Hence, the primary purpose of this study is to characterize adaptation and recovery from adaptation in response to pulsatile stimulation in the auditory nerve of pediatric and adult cochlear implant users. A secondary purpose of this study was to examine if the inter-subject variability in adaptation that is measured in a multi-pulse paradigm is related to the variability in temporal interactions, especially refractoriness, in a two pulse paradigm. Finally, given that the peri- and post-stimulatory effects of adaptation and refractoriness will vary with stimulus level due to changes in both, the recruited nerve fiber population and the temporal interactions at the neural membrane of individual fibers, these phenomena are examined at various combinations of the masker and probe level. Since the most prominent modulations in the speech signal occur at a rate of 3-4 Hz (Houtgast & Steeneken, 1985) due to syllabification, groups of nerve fibers may be required to encode onsets at intervals of 250 to 350 ms. Within each modulation, the adapting signal, referred to as the masker may be on for a part of the duration

23 13 allowing nerve fibers to recover from their responsiveness during the remainder of the time. Hence, the adapting stimuli used in this study are brief (100 ms) constant amplitude pulse trains delivered at a rate of 1800 pps on a single basal (fifth) electrode. Adaptation or, more specifically, the decrement in the amplitude of the neural response is examined at the end of the 100 ms masker pulse train and the recovery of the neural response to a probe stimulus is examined at approximately logarithmically spaced time intervals up to 250 ms after the cessation of the adapting stimulus. The selected rate of 1800 pps is within the capabilities of current speech processors and may likely be delivered at levels that would produce a robust neural response without causing discomfort. It is acknowledged that speech is characterized by a spectral alternation in the time domain and hence, we must be concerned with temporal interactions across spectral channels rather than simply within a channel. However, the stimulus paradigm used here may be considered a simulation of channel interaction if the masker and probe, although delivered to the same electrode, are thought of as attenuated versions of stimuli delivered to different electrodes. Additionally, it is recognized that the adapting stimulus in real listening situations would be itself modulated and likely produce less peri- and post-stimulatory adaptation than will be measured in this study. Also, the term masker may be a misnomer as the preceding stimulus may potentially enhance, rather than reduce, the response to a succeeding probe. The neural response used to characterize temporal interactions is the electrically evoked compound potential (ECAP). The ECAP is a summed electrical response of an ensemble of fibers to electrical stimulation (Miller, Brown, Abbas & Chi, 2008) such as delivered by an electrode (pair) in the implanted array. Given that individual fibers

24 14 respond to onsets of speech segments with high firing probability, electrical stimulation may ensure that onsets are encoded by the synchronous firing of a large number of nerve fibers and consequently, the ECAP is likely a suitable measure of onset encoding. The contribution of individual fibers to the ECAP may be uniform for a distant recording electrode (Abbas & Miller, 2004) but is biased significantly towards fibers closest to a proximal recording electrode such as used in this study (Abbas, Brown, Shallop, Firszt, Hughes, Hong & Staller, 1999). The response in cochlear implant users may be recorded, as in this study, using an implanted electrode (Abbas et al, 1999). An intracochlearly recorded ECAP is characterized by a negative peak (N1) that occurs about 0.2 to 0.4 ms after the onset of a stimulus pulse and a smaller positive peak (P2) that follows (Abbas et al, 1999). The amplitude of the ECAP is the voltage difference between P2 and N1. The ECAP amplitude and waveform are dependent on, both, the number and synchrony of action potentials. The index used to characterize the temporal interactions at the neural membrane that result from a the presentation of masker pulse/pulse train and a probe pulse in rapid succession is the ratio between the ECAP amplitude produced by the probe with a preceding masker and the ECAP amplitude produced by the same probe when presented singly. It is recognized that decrements in ECAP amplitude due to a change in the site of excitation (Stypulkowski & van den Honert, 1984), reduction in action potential amplitude (Zhang et al, 2007) or increased temporal jitter (Litvak et al, 2001; Litvak et al, 2003a; Wilson et al, 1997a, b) cannot be differentiated from rate adaptation, i.e., decrease in the neural firing rate.

25 Specific research questions and hypotheses The specific research questions and hypotheses are as follows: i. What are the effects of stimulus parameters, viz., masker level, probe level and masker-probe interval (MPI), on the amplitude of the probe evoked ECAP when the masker is a single electric pulse? Do the stimulus effects vary across the age groups? It is hypothesized that for a fixed masker level, the normalized ECAP amplitude will be 1 for all probes higher than the masker and will increase with probe level for all probes lower than the masker. For a fixed probe level, the normalized ECAP amplitude will be 1 for all maskers lower than the probe and less than 1 for all maskers higher than the probe. However, for any given probe level, the effectiveness of a masker is not expected to increase as masker level is progressively increased above the probe level. The normalized ECAP will be smaller at the shorter MPI (1.2 ms) as compared to the longer MPI (2 ms) used in this study and this will be true for all masker probe combinations that produce normalized ECAP amplitudes of less than 1. For any maskerprobe level combination, the normalized ECAP amplitudes at either MPI are expected to be smaller in children than adults. ii. What are the effects of stimulus level and age on adaptation? Specifically, does the normalized (asymptotic) ECAP amplitude at the end of a 100 ms pulse train vary across stimulus level and age? It is hypothesized that children will show greater adaptation, i.e., lower normalized ECAP amplitudes near the end of the pulse train as compared to adults. The effects of stimulus level must be explored and cannot be easily predicted. It is possible that high stimulus levels may produce greater driven activity at the onset of the pulse

26 16 train and consequently, greater activity dependent adaptation and lower normalized ECAP amplitudes at asymptote in comparison to a low level stimulus which produces less onset activity and less rate adaptation in the excited fiber population. Conversely, it is also possible that the higher level stimulus may be able to drive nerve fibers that are in an adapted state to fire synchronously while a low level stimulus is unable to do so; consequently, the higher level pulse trains may be associated with higher normalized ECAP amplitudes than the lower level pulse trains. iii. What are the effects of the stimulus factors (pulse train masker level, probe level and masker probe interval) and age on recovery from adaptation? It is hypothesized that recovery will be faster in adult subjects as compared to children. Again, the effects of stimulus level cannot be predicted easily. To the extent that the decrement in ECAP amplitude is simply the result of stochastic independence across nerve fibers, recovery of the normalized ECAP amplitude requires the recovery from refractoriness and the restoration of synchrony and hence, may be rapid. However, if ECAP amplitude reflects a decrease in firing probability such as resulting from changes in the electrical status of the neural membrane and/or chemical balance between the intraand extra-cellular spaces, then it is likely that recovery may be slow. At the very least, it may be predicted, that for a given level of the masker pulse train, higher level probes will produce faster recovery because they recruit fibers that were not within the excitation field of the masker and/or because they are able to drive fibers in varying states of adaptation/recovery unlike lower level probes. Also, for any given level of the single pulse probe, the effectiveness of lower level masker pulse trains may be limited but not necessarily negligible given that sub-threshold adaptation has been observed. However,

27 17 unlike refractoriness, adaptation by increasingly higher levels of the masker could produce slower recovery if higher initial driven rates are responsible for greater adaptation. iv. Can we predict the adaptation and recovery for a masker pulse train-probe pulse combination from the two pulse interaction between a single masker probe pulse combination at the same levels? It is hypothesized that although recovery from refractoriness and adaptation may involve different phenomena, the two maybe positively correlated so that the normalized ECAP amplitude obtained in a two pulse paradigm for any level combination of the masker and probe will be predictive of the adaptation and recovery of the normalized ECAP amplitude obtained by using a masker pulse train and a probe pulse at those levels.

28 18 CHAPTER 2 REVIEW OF LITERATURE Most cochlear implants code speech as trains of biphasic electric pulses that are amplitude modulated by the temporal envelope of the speech signal. Specifically the signal is passed through a set of band-pass filters and the temporal envelopes of the output of the various filters are used to modulate the amplitude of electric pulses delivered to the different electrodes. It is hoped that the auditory nerve fibers will, as a group, encode the modulation waveform. It is not surprising that the response to each individual pulse is dependent not simply upon the characteristics, viz., level, shape, duration of that pulse but also upon the context in which the pulse is presented. In particular, the past history of stimulation, and response, affects the response to any given pulse. Given that cochlear implants today, typically, use sequential pulsatile stimulation, these interactions are not spatial interactions of the electric fields but, rather, are attributable to temporal interactions at the neural membrane. 2.1 Two Pulse Paradigms Temporal interactions have been studied with two-pulse paradigms (Cartee, Miller & van den Honert, 2006; Cartee, van den Honert, Finley & Miller, 2000; Dynes, 1995; Finley, Wilson, van den Honert & Lawson, 1997; Grill & Mortimer, 1995; Morsnowski, Charasse, Collet, Killian & Muller-Deile, 2006; Shepherd & Javel, 1997, 1999; Shepherd, Roberts & Paolini, 2004; Stypulkowski and van den Honert, 1984; van den Honert & Mortimer, 1979). Stimulus paradigms that involve the study of the effect of either phase of a biphasic pulse on the response to the other phase constitute a special

29 19 case (Shepherd & Javel, 1999; Grill & Mortimer, 1979; van den Honert & Mortimer, 1979) as each phase may be considered a monophasic pulse of the two pulse paradigm Integration Current integration between pulses and/or phases of a single pulse is known to occur at the neural membrane. The delivery of the non-excitatory phase in rapid succession after the excitatory phase of a biphasic electric pulse may abolish the action potential while an increase in the interphase gap may allow the probability of spike occurrence to approach that of an excitatory monophasic pulse (van den Honert & Mortimer, 1979; Shepherd & Javel, 1999). Similarly, the spike threshold for the initial excitatory phase of a biphasic charge-balanced electric pulse is lowered by slower charge delivery during a prolonged lagging non-excitatory phase (Shepherd & Javel, 1999). The initial delivery of the hyperpolarizing or non-excitatory phase is also known to raise the spike threshold of the subsequent excitatory phase (Shepherd & Javel, 1999). Depolarizing pulses that are subthreshold or associated with a less than 100% firing efficiency may lower the threshold for a subsequent pulse. In single fiber studies, using pulse pairs of equal amplitude (Cartee et al, 2000; Cartee et al, 2006) and pulse pairs with a variable amplitude of the second, i.e., probe, pulse (Dynes, 1995), it has been observed that a pulse delivered well below resting threshold is able to sensitize the neural membrane so that the spike threshold for a subsequent pulse is reduced. Enhancement of the ECAP amplitude using similar stimulus paradigms in human subjects (Abbas, Etler, O Brien & Brown, 2009) and the observation of non-monotonic recovery of the ECAP in animals (Stypulkowski & van den Honert, 1984) and in humans (Finley et al, 1997) have also been attributed to the same or similar phenomena, variously

30 20 referred to as summation (Cartee et al, 2000; Cartee et al, 2006), sensitization (Dynes, 1995), conditioning (Abbas et al, 2009) and integration (Finley et al, 1997; Stypulkowski & van den Honert, 1984). The amount of the initial sensitization increases as the level of the first pulse, i.e., the prepulse or conditioning pulse, approaches resting threshold (Dynes, 1995) possibly because the neural membrane is almost depolarized to threshold by the first pulse with the second or probe pulse simply serving to trigger the action potential (Abbas, Personal Communication). The period of this sensitization appears to extend for about µs (Cartee et al, 2000, Stypulkowski & van den Honert, 1984) or up to 500 µs and nearly a millisecond (Dynes, 1995; Finley et al, 1997) following the prepulse. The summation time constants have been found to vary from 147 µs for intra-meatal stimulation (Cartee et al, 2000) to 400 µs for scala tympani stimulation (Cartee et al, 2006) and thought to reflect differences in membrane capacitance along the fiber length. Specifically, Cartee et al (2006) based on their previous modeling work consider the shorter time constants to be consistent with stimulation on the distal nodes of the peripheral processes or on the central axon while they consider the longer time constants to be consistent with stimulation on a peripheral node adjacent to the large capacitance of the cell soma. Stypulkowski & van den Honert (1984) found stimulus integration to be associated exclusively with the N1 peak of the ECAP and hence, attributed the phenomenon to the longer time constants of the unmyelinated peripheral processes. They did, however, report integration of spikes at the cell soma so that a single spike propagated beyond. Given that this phenomenon has been observed in the ECAP of long-term deafened human subjects who likely do not have peripheral processes, stimulus integration is likely

31 21 located to other parts of the ANF as well. Specifically, demyelination of the cell soma and central axon, in human subjects could hypothetically prolong the time constants and produce integrative phenomena that may not always be evident in acutely deafened or normal control animals. In addition to the enhancement of the response to subsequent stimuli, the relative spread may be increased for a period of 1 ms following a subthreshold pulse (Dynes, 1995). Also, a period of desensitization lasting 1 to 5 ms may follow the period of enhancement with threshold elevations of 1-2 db having being observed in cat ANFs during this period (Dynes, 1995). The integrative phenomena have been attributed to the charge holding capacity of neural membrane acting as a parallel RC circuit (Cartee et al, 2000, 2006) and/or the nonlinear processes at the neural membrane (Dynes, 1995; Finley et al, 1997). The transient increase in excitability produced by a prolonged pulse that itself hyperpolarizes the membrane and, conversely, the transient decrease in excitability produced by a prolonged subthreshold pulse that itself depolarizes the membrane have been considered to be consistent with changes in membrane resistance or, more specifically, the gating mechanisms of the Na + channels (Grill & Mortimer, 1995). Similarly, the observation that integrative interactions occur between two biphasic pulses would also suggest the occurrence of non-linear mechanisms. The explanation offered by Dynes (1995) is considered here. The subthreshold conditioning pulse likely depolarizes the membrane so that Na + activation gates open to allow Na + influx and further depolarization. While the closure of inactivation gates and K + leakage would repolarize the membrane, both occur with a time lag. The membrane depolarization and decreased resistance consititue

32 22 a lowering of spike threshold so that a subsequent pulse may easily trigger a spike. Desensitization involves K + outflow and closure of the Na + channel inactivation gates Refractoriness Just as with other excitable membranes, the occurrence of a spike in the ANF is followed by a brief refractory period during which the membrane restores its resting potential. The early period following the spike forms the absolute refractory period during which the ANF will be unable to fire another action potential regardless of the intensity of the stimulus. The absolute refractory period is followed by a relative refractory period during which the ANF is able to produce a spike but at an elevated threshold. The pulse that evokes the spike and puts the ANF in a refractory state is typically referred to as the masker pulse while the subsequent pulse that is used to assess the state of refractoriness is referred to as the probe. It is noted that a so-called masker pulse may have masking effects on the probe response in certain fibers and conditioning effects in other fibers. The ECAP amplitude may be examined as a function of the masker-probe interval (MPI) to delineate recovery of the fiber population. Yet, it is noted that the ECAP recovery function may not be predictive of the individual fiber recovery patterns. Individual fiber recovery will depend upon the effective stimulus at the neural membrane (Miller, Abbas & Robinson, 2001) as well as the membrane properties both of which may vary across the fiber population. Furthermore, given that the ECAP amplitude is also dependent upon the individual spike amplitudes as measured at the recording electrode and the timing synchrony of action potentials across the responding population, characterization of refractory effects with the ECAP may not yield an accurate estimate

33 23 of the neural output (Abbas & Miller, 2004). Single fiber studies have limitations in that they may not sample the population adequately (Miller et al, 2001). The absolute refractory period in cat ANFs has been reported to be greater than or equal to 500 µs and, typically, ranges from 600 to 700 µs (Dynes, 1995). Other studies (Miller et al, 2001; Stypulkowski & van den Honert, 1984) with the same animal model have yielded lower estimates, i.e., 300 to 350 µs. ECAP measures in human subjects have yielded a median value of ~ 400 µs for the absolute refractory period (Morsnowski et al, 2006). Also, Shepherd et al (2004) have observed a significant positive relationship between the duration of deafness and absolute refractory periods in rat ANFs. Single fiber data suggest that threshold, if defined as a 50% firing efficiency, is more or less restored by 3 ms after the masker pulse but not if more stringent criteria are used to define threshold (Miller et al, 2001). The N1 response of the ECAP, thought to originate from the peripheral processes, and the N0 response, thought to originate from the central axon, both showed a complete recovery by 4 ms in the data presented by Stypulkowski and van den Honert (1984). ECAP recovery appears to be complete in cats by 4 ms post-stimulation but not in guinea pigs (Matsuoka, Abbas, Rubinstein & Miller, 2000). Recovery of the ECAP appears to be complete by about 5 ms in the limited human data reported by Finley et al (1997). Data from larger groups of implant users suggest that recovery following stimulation in monopolar mode is largely complete by about 5 ms (Brown, Abbas, Borland & Bertschy, 1996) although considerable variability was evident across subjects and electrodes within subject (Brown et al, 1996; Brown, Abbas & Gantz, 1998). Recently, it has been suggested (Botros & Psarros, 2010) that slow recovery of the ECAP is consistent with the recruitment of a large neural population

34 24 at a lower firing efficiency. This would seem plausible given that recovery in single fibers is known to be level dependent (Miller et al, 2001). Mean time constants of recovery have been variously reported as 2 ms (Dynes, 1995), ~ 400 µs (Miller et al, 2001), about 550 µs for fibers thought to be stimulated centrally and 838 µs for fibers likely to be stimulated peripherally (Cartee et al, 2006). Reanalysis of Dynes data (1995) by Cartee et al (2000) using their equation revealed time constants of about 1.5 ms. Human ECAP data have yielded a median time constant of recovery that is 425 µs (Morsnowski et al, 2006) and mean time constants of 1.5 to 2.5 ms for various etiologies and durations of deafness (Gantz, Brown & Abbas, 1994). Estimates of the absolute refractory periods and time constants of recovery across studies vary depending on several factors. The estimates vary considerably depending on the shortest MPI studied (Miller et al, 2001), i.e., unless the initial part of the recovery function is adequately defined, absolute refractory period and the time constant of recovery may be spuriously altered depending upon the equations used to quantify recovery. Further, as different levels of a single pulse may produce spikes at different sites along the ANF (Matsuoka et al, 2000; Stypulkowski & van den Honert, 1984), the level and polarity dependence of refractory time constants have been thought to reflect the properties of different sites of excitation even in the same species (Matsuoka et al, 2000). Additionally, the site of spike generation may potentially differ for the masker and probe pulse (Stypulkowski & van den Honert, 1984). For this and other reasons, recovery time estimates discussed earlier depend upon the spike generation and spike conduction (Cartee et al, 2006; Stypulkowski & van den Honert, 1984) to the recording site that has been typically located on the nerve trunk in animal studies of single fiber or