Sound Processing Based on Neural Models
Contact
Background
Cochlear implant users have difficulty hearing speech in background noise and also have very poor perception of temporally complex sounds including music and tonal languages such as Chinese (Rubinstein, 2003; McDermott, 2004). These limitations can be conceived of as due to limitations in information transfer that may in part be overcome through improving the speech processing strategies used in cochlear implants The proposed research builds upon recent work in our laboratories where we have developed a model of the auditory nerve to electrical stimulation that accurately predicts the response of auditory nerve fibres to electrical stimulation derived from one channel of a cochlear implant speech processor and delivered to one electrode of the cochlear implant. We hope to utilise this model in cochlea implants as at present, no knowledge of how neurons respond to electrical stimulation is used in cochlear implants, even though the neurons are the very target of the stimulation.
Research Aims
Fine Temporal Structure. In normal hearing individuals, high frequency fluctuations (temporal fine structure) of speech or music is encoded in the timing between the action potentials of auditory neurons. However, at present cochlear implants only use the basic waveform envelope in stimulation strategies. We will test if electrical stimuli carrying fine temporal structure can generate a better representation of fine temporal structure in the neural response. Experimentally, the first acoustic stimuli tested will be pure tones delivered to the cochlea. Later we will deliver more complex stimuli containing fine temporal stimuli structure including the response to more complex sounds, and a full speech spectrum.
Multiple electrodes. At present, the testing of the neural model in the guinea pig has used a cochlear implant electrode that only stimulates on 1-3 electrodes, rather than 22 electrodes used in the human implant. We will employ electrodes with up to 8 active bands (already in use) and we aim to develop equipment to allow us to stimulate near-simultaneously on multiple electrodes to more closely emulate cochlear implant stimulation.
Population response. Our present model only supplies the output of a single auditory nerve fibre. We will expand this into a population model (Bruce et al., 2000) by including the range of thresholds to electrical stimulation seen in auditory neruons.
The outcome of this proposed research will determine if the use of neural modeling in cochlear implants can increase or maintain the output of information that is carried within auditory nerve fibres and benefit patient hearing. The outcome has significant clinical relevance for all types of hearing loss. We anticipate testing the outcomes of these models in the near future in patients that have implants with a percutaneous plug. The neural modelling could benefit patients with neural prostheses other than the cochlear implant as no other group has previously modelled the response to patterned electrical stimulation of nerves in any part of the body.
Outcome
If the incorporation of neural models in implants is successful, improved music perception would benefit all cochlear implant patients and improvements to tonal language perception would open a large market for cochlear implants in many Asian countries that is largely unaddressed at present.
The existing neural model that we have developed does not require exceptionally detailed calculations and the model is fast enough to be utilised in real time within a cochlear implant. In its present form the model is unlikely to require further modifications to cochlear implant hardware, but rather be implemented via firmware on existing cochlear implant DSP chips.
Our group is unique in that we have the expertise in single unit electrophysiology of auditory nerves, mathematically modeling of cochlea and neural responses and auditory signal processing to approach this problem. Although other groups have previously produced mathematical models of the cochlear approach to electrical stimulation, no other groups have tested or verified these models within invivo models. Our group has already established this method of validation.
Techniques
- Electrophysiology - extracellular neuron recordings
- Computational Neural Modelling
Staff
- Professor Stephen O'Leary (Academic Group Leader)
- Mr. Leon Heffer (Research Fellow, Physiology)
- Ian Jakovenko (Mathematician)
- Dr David Sly (Research Fellow, Physiology)
- Ms. Ricki Minter (Research Assisstant, Physiology)
Support Staff
- Mr. Rodney Millard (Engineering)
Collaborators
- Professor Peter Seligman
- Professor Hugh McDermott
- Professor Chris Van Den Honert
- Mr. Luke Campbell (Medical Student; Physiology)
Funding
Recent Publications
- Campbell, L.J., Sly, D.J., O'Leary, S.J. Prediction and Control of the Stochastic Neuronal Response to Pulsatile Electrical Stimulation. Submitted European Journal of Neuroscience, December 2008.
- Cambell, L.J. Sly, D.J., O'Leary, S.J. An electrically-stimulated auditory nerve model: Testing with variable amplitude pulse trains derived from speech. Poster presentation, World Congress of Neuroscience Meeting, International Brain Research Organisation, Melbourne, July, 2007.
- Campbell, L.J. Advance Medical Science Thesis. Computational Modelling of Auditory Neuron Responses to Pulsatile Electrical Stimulation. (Supervisors: Stephen O'Leary and David Sly). Completed May 2007.
- Sly, D.J., Heffer, L.F., White, M.W., Shepherd, R.K., Birch, M.J., Minter, R.M., Nelson, N.E., Wise, A.W. and O'Leary, S.J. (2007) Deafness alters auditory nerve fiber responses to cochlear implant stimulation. European Journal of Neuroscience, 26, 510-522.
- McDermott, H.J. (2004) Music perception with cochlear implants: a review. Trends Amplif, 8, 49-82.
- Rubinstein, J. & Hong, R. (2003) Signal coding in cochlear implants: exploiting stochastic effects of electrical stimulation The Annals of otology, rhinology & laryngology Supplement, pp. 14-19.
- Bruce IC, Irlicht LS, White MW, O'Leary SJ, Clark GM. (2000) Renewal-process approximation of a stochastic threshold model for electrical neural stimulation. J Comput Neurosci. 9(2):119-32.
- Bruce IC, White MW, Irlicht LS, O'Leary SJ, Clark GM. (1999) The effects of stochastic neural activity in a model predicting intensity perception with cochlear implants: low-rate stimulation. IEEE Trans Biomed Eng. 46(12):1393-404.
- Bruce IC, Irlicht LS, White MW, O'Leary SJ, Dynes S, Javel E, Clark GM. (1999) A stochastic model of the electrically stimulated auditory nerve: pulse-train response. IEEE Trans Biomed Eng. 46(6):630-7.
- Bruce IC, White MW, Irlicht LS, O'Leary SJ, Dynes S, Javel E, Clark GM. (1999) A stochastic model of the electrically stimulated auditory nerve: single-pulse response. IEEE Trans Biomed Eng. 46(6):617-29.