Pseudospontaneous activity: stochastic independence of auditory nerve fibers with electrical stimulation

Abstract

We describe a novel signal processing strategy for cochlear implants designed to emphasize stochastic independence across the excited neural population. The strategy is based on the observation that high rate pulse trains may produce random spike patterns in auditory nerve fibers that are statistically similar to those produced by spontaneous activity in the normal cochlea. We call this activity `pseudospontaneous'. A supercomputer-based computational model of a population of auditory nerve fibers suggests that different average rates of pseudospontaneous activity can be created by varying the stimulus current of a fixed-amplitude, high-rate pulse train, e.g. 5000 pps. Electrically-evoked compound action potentials recorded in a human cochlear implant subject are consistent with the hypothesis that such a stimulus can desynchronize the fiber population. This desynchronization may enhance neural representation of temporal detail and dynamic range with a cochlear implant and eliminate a major difference between acoustic and electric hearing.

Introduction

Temporal properties of auditory nerve fiber responses are substantially different when the fibers are stimulated electrically with a cochlear implant than with acoustic stimulation in the normal hearing ear. Electrical stimulation produces highly synchronous activity across the excited neural population (Moxon, 1971). Acoustic stimulation produces much less across-fiber synchrony and much more within-fiber jitter due to the stochastic properties of the inner hair cell synapse (Kiang and Moxon, 1972). It seems reasonable to expect that if a more physiologic pattern of temporal responses could be obtained from electrically stimulated auditory neurons, better hearing would result with a cochlear implant.

A major difference between the deaf and hearing ears is the relative absence of spontaneous activity in the deafened cochlea (Kiang et al., 1970; Liberman and Dodds, 1984). This has implicated the inner hair cell synapse as the source of spontaneous activity (Sewell, 1984) although recently it has been suggested that elimination of spontaneous activity with ototoxic deafening is not as complete as might be suspected (Shepherd and Javel, 1997). We expect that restoration of physiologic levels of spontaneous activity to the deafened cochlea would be a productive approach to speech processor design for a number of reasons:

• The normal auditory nerve is spontaneously active in quiet (Liberman, 1978). Sound produces a slowly progressive within and across fiber synchronization as intensity is increased (Schoonhoven et al., 1997). Replication of this phenomenon should allow greater dynamic range and more orderly loudness growth.

• Studies of `stochastic resonance' demonstrate increased temporal resolution in sensory systems when independent noise is present in a set of parallel detectors (Collins et al., 1995; Parnas, 1996). The spiral ganglion is quite comparable to such parallel detectors and spontaneous activity in each fiber is a form of independent noise (Johnson and Kiang, 1976).

• Loss of spontaneous activity is one proposed mechanism for tinnitus (Kiang et al., 1970) and its restoration may potentially improve tinnitus suppression by cochlear implants.

We have investigated a number of possible approaches to increasing stochastic independence in responses of auditory nerve fibers to electrical stimulation. The first is to employ the pulse rate that maximizes the stochastic properties of the refractory period. We have demonstrated in models and measures of the electrically-evoked compound action potential (EAP) that an interpulse interval between 0.9 and 1.0 ms substantially decreases the slope of the neural input-output function and EAP growth function (Matsuoka et al., 1997). This is due to a dramatic increase in the noise of the voltage-sensitive sodium channel during the relative refractory period (Rubinstein et al., 1997). While promising, this interesting effect is exquisitely sensitive to the interpulse interval and has so far only been studied for two pulse pairs. Any channel interactions present in a cochlear implant could make it difficult or impossible to control the interpulse interval `seen' by a given neural population.

A second approach is to modulate high-rate pulse trains with the envelope of the filtered, compressed speech signal. At low rates, amplitude of the EAP to successive pulses in the train show an alternating pattern – suggesting refractory effects which are evident due to the high degree of synchronization (Abbas et al., 1997; Wilson, 1997; Wilson et al., 1994, Wilson et al., 1997a, Wilson et al., 1997b, Wilson et al., 1997c). At high rates, above 2 kHz in humans, the response amplitudes are constant after the first few pulses consistent with an increased stochastic independence of the firing patterns of the fiber population (Wilson, 1997; Wilson et al., 1997a, Wilson et al., 1997b, Wilson et al., 1997c). This approach requires modulation of high rate stimuli and produces small compound action potentials suggesting that small numbers of fibers are available (not absolutely refractory) at any time. More faithful representation of the temporal details of the speech envelope might be achieved if it is possible to maintain high levels of stochastic independence yet keep a larger pool of neurons out of the absolute refractory period.

A third approach, the topic of this work, is to recreate if possible the independent noise sources found in normal auditory nerve fibers. We have observed that spike trains can be produced in simulated electrically-activated auditory nerve fibers that have statistical properties similar to spontaneous activity in normal spiral ganglion cells. We call this `pseudospontaneous' activity. In this paper we present our preliminary description of this activity in simulated auditory nerve fibers as well as recordings of the electrically-evoked compound action potential in a human implant subject which support the existence of the phenomenon.