Unraveling the electrically evoked compound action potential

https://doi.org/10.1016/j.heares.2005.03.020Get rights and content

Abstract

With the advent of eCAP recording tools such as NRT and NRI for cochlear implants, neural monitoring has become widely used to ascertain the integrity of the neural/electrode interface as well as for assisting in the setting of program levels. The basic concepts of eCAP recordings are deduced from the acoustical equivalent of the electrocochleogram. There are, however, indications that under electrical stimulation some of these do not hold, like the unitary response concept (i.e., the principle that every fiber produces the same contribution to the eCAP). Computer modeling has proven to be a valuable tool for gaining insight into the functioning of electrical stimulation. In this study the extension of a three-dimensional human cochlea, incorporating back-measuring capabilities, is described. Using this new model, the contribution of single fiber action potentials (SFAPs) to the measured eCAP is investigated. The model predicts that contrary to common belief – the compound action potential as measured by the cochlear implant system does not necessarily reflect the propagated action potential along the auditory nerve.

Introduction

The recording of electrically evoked compound action potentials (eCAPs) of the auditory nerve has become widespread since the introduction of neural response telemetry (NRT) by Cochlear Ltd. (Sydney, Australia) and neural response imaging (NRI) by the Advanced Bionics corporation (Sylmar, CA, USA). These systems allow easy acquisition of the eCAP through the cochlear implant system without the need of extra recording or stimulating electrodes such as a trans-tympanic needle. The fundamentals of auditory CAP recordings are known from the acoustically evoked equivalent, which is the electrocochleogram. The latter is used in clinical practice as a reliable way to measure a frequency specific objective audiogram (Schoonhoven et al., 1996, Schoonhoven et al., 1999). Stimuli with alternating polarity are used in such acoustically evoked recordings to remove the cochlear microphonic (CM), which results from outer hair cell responses. One of the mainstays of these recordings is that the CAP response can be described as a superposition of unitary responses, meaning that each nerve fiber contributes equally to the signal (Versnel et al., 1992, Goldstein and Kiang, 1958), and that the amplitude of the recorded fiber response correlates with the number of excited nerve fibers. The recording of eCAPs gives rise to some specific issues compared to the electrocochleogram, e.g., now the electrical artifact has to be suppressed instead of the CM. The question that remains is whether all the other principles of acoustical CAP recordings are applicable for eCAP recordings. Moreover, to date the clinical value of using eCAP input/output functions and thresholds to set processor parameters remains limited to finding contours for the levels in the programs. This objective data have to be supplemented by behavioral data to get functional programs (Abbas et al., 1999, Seyle and Brown, 2002, Smoorenburg et al., 2002). It remains unclear, however, what the fundamental problems are why it is not possible to build programs that patients prefer based only on eCAP. Additionally, the ability to measure refractory properties (Miller et al., 2000) and the possibility to obtain objective measures of spatial selectivity (Cohen et al., 2003, Abbas and Brown, 2000, Frijns et al., 2002) are not yet sufficiently validated in terms of their clinical applicability. More specific questions that need to be answered are, to what extent these objective measures are applicable to the fitting of children or can be used to assess neural degeneration in regions of the cochlea.

To gain further insight into the working mechanisms of electrical stimulation of the auditory nerve by a cochlear implant, a detailed computer model of the cochlea has been developed at the Leiden University Medical Center (Frijns et al., 2000, Briaire and Frijns, 2000a). The model consists of two parts, a 3D volume conduction model and an active auditory nerve fiber model. The volume conduction part provides insight into the distribution of the current through the cochlea (Briaire and Frijns, 2000b) and how this distribution can be influenced by, for example, electrode orientation. In and of itself however, the potential distribution does not tell which fibers will react and how the stimulus waveform affects the response. For this purpose the nerve fiber model is used. Using the potential distribution along the nerve fibers for its input, this model calculates which fibers and what part of these fibers will be excited (Frijns and ten Kate, 1994; Frijns et al., 1995). Initially, both models were based upon the cochlea of a guinea pig, our experimental animal, with corresponding fiber morphology and nodal kinetics. The presence of cross-turn stimulation, the unintended and unwanted activation of nerve fibers from a higher turn than where the stimulating electrode is located, and the influence of the electrode position were the main outcomes of this initial model (Frijns et al., 1995, Frijns et al., 1996a). To generate an applicable human model, the animal volume conduction model was extended to use realistic electrode geometries and to match the human cochlear anatomy. This also gave us a tool to investigate the validity of transferring results obtained from animal experiments to the human situation (Frijns et al., 2001). Recently, one of the main outcomes of this study, a decrease in threshold and an increase in neural excitation at a fixed current level when the electrode is moved to a more peri-modiolar position, was confirmed with intra-operative electrically evoked auditory brainstem response (EABR) measurements with the HiFocus electrode in lateral and medial positions by Firszt et al. (2003).

Previous studies, with integrated use of neural and volume conduction models from a number of groups (Hanekom, 2001, Rattay et al., 2001a), typically focused on the so-called forward problem, i.e., predicting the neural excitation pattern. To our knowledge, the first attempt to simulate eCAPs was a preliminary study with a guinea pig model that was in itself just capable of solving the forward problem (Frijns et al., 1996b). It used the assumption that the unitary response concept was also valid for the electrically stimulated cochlea and led to the conclusion that just the site of excitation along the auditory nerve fiber cannot account for the latency differences observed. To answer more subtle questions about the value and possibilities of eCAP recordings and the validity of the unitary response concept a more sophisticated model is needed, which is able to solve the full backward problem, i.e., the calculation of the eCAP response as recorded via intra-cochlear electrode contacts.

As stated above, the geometry of the cochlea and the implant used in our previous studies were already in line with the human anatomy. However, the kinetics and morphology of the nerve fibers were still based on the guinea pig, which likely compromised the applicability of the findings to humans. The main anatomical difference between the guinea pig and human primary auditory nerve fiber is the morphology of its cell body, which is unmyelinated in humans as contrasted with all other mammals. From Rattay’s model study (Rattay et al., 2001b) it is known that the unmyelinated cell body in the human fiber is likely to play an important functional role, as it induces a delay in the conduction of the action potential (AP) along the fiber. Also the length of the peripheral process in humans is much longer than in guinea pigs. Therefore, there are presumably more than three (the value used in the guinea pig models) inter-nodal segments in the human peripheral process, although there is no formal histological evidence available.

The present study aims at deriving a fundamental understanding of the processes underlying eCAP recordings in humans, both in terms of the contributions of the individual nerve fibers to the overall signal as well as to what extent this signal yields clinically relevant information about functional aspects of electrical stimulation. For this purpose the neural model has been extended to incorporate an unmyelinated cell body and an unmyelinated pre-somatic region. In addition, the algorithms of the volume conduction model have been upgraded to calculate the eCAP from single fiber responses in order to generate simulated wave forms that can be compared with actual recordings.

Section snippets

The forward problem: simulating neural excitation in the human cochlea

The computational model of the electrically stimulated human cochlea as described in Frijns et al. (2001) forms the basis of the computational model used in the present study. The model geometry is a realistic three-dimensional representation of the human cochlea with a model representation of the Clarion HiFocus cochlear implant (Figs. 1(a)–(c)). Details of this geometry have been published in the aforementioned paper. Unless stated otherwise, the implant array is in a peri-modiolar position

Results

The most detailed presentation of the outcome of the forward problem is the so-called excitation profile as shown in Fig. 4. Such a plot depicts which nerve fibers get excited at various stimulus levels. The location of the initial excitation (peripheral process, cell body or modiolar axon) is indicated by the degree of shading. Fig. 4(a) shows such an excitation profile for the standard conditions as defined in Section 2, as calculated with the old, guinea pig based, fiber (gSEF) used in

Discussion

In this study the mechanisms behind the eCAP are investigated with a computational model. The model consists of a realistic three-dimensional cochlea model combined with an active, non-linear nerve fiber model. To be able to calculate the eCAP properly, the model had to be extended to allow for recording of the currents generated in the nerve fiber model through the volume conduction model of the cochlea. The generated currents and the recording algorithm used in a long and homogeneous nerve

Acknowledgments

This research was financially supported by grants from the Hoogenboom-Beck-Fund and the Heinsius Houbolt Fund. We thank Prof. Dr. J.J. Grote, former head of our department, for his continuing support.

References (36)

  • F. Rattay et al.

    A model of the electrically excited human cochlear neuron. ii. Influance of the three-dimensional cochlear structure on neural excitability

    Hear. Res.

    (2001)
  • F. Rattay et al.

    A model of the electrically excited human cochlear neuron. i. Contribution of neural substructures to the generation and propagation of spikes

    Hear. Res.

    (2001)
  • H. Versnel et al.

    Round-window recorded potential of single-fibre discharge (unit response) in normal and noise-damaged cochleas

    Hear. Res.

    (1992)
  • P.J. Abbas et al.

    Electrophysiology and device telemetry

  • P.J. Abbas et al.

    Summary of results using the nucleus ci24m implant to record the electrical evoked compound action potential

    Ear Hear.

    (1999)
  • J.B. Firszt et al.

    Electrically evoked auditory brain stem responses for lateral and medial placement of the clarion hifocus electrode

    Ear Hear.

    (2003)
  • J.H. Frijns et al.

    Initial evaluation of the clarion cii cochlear implant: speech perception and neural response imaging

    Ear Hear.

    (2002)
  • Frijns, J.H.M., Briaire, J.J. , 1999. The consequences of species differences in cochlear anatomy on the portability of...
  • Cited by (81)

    View all citing articles on Scopus
    View full text