At the University of Iowa, Dr. Jay Rubinstein, a computational scientist and assistant professor of otolaryngology, over the past ten years has been developing a computer model to investigate the biophysics of cochlear implants. The simulation runs on the CRAY C90 at the San Diego Supercomputer Center (SDSC) and represents the primary auditory neurons of both a cat and guinea pig. The auditory performance in both intact and deafened animals with a cochlear implant is being compared via electrical stimulation of the auditory system. The ultimate goal has been focused on the major improvement of information transfer for human patients with all types of auditory prostheses implants.
Since many years, the research on auditory prostheses implants in the United States is supported by the Neural Prosthesis Program (NPP), as initiated by the National Institutes of Health (NIH), and the National Institute on Deafness and Other Communication Disorders (NIDCD). The speech processor forms an essential element, common to all people with cochlear or cochlear nucleus implants, since this device converts the wide bandwidth electrical signal from a microphone to a range of signals stimulating the implant user to recognize both speech and environmental sounds. Persons with severe sensorineural hearing loss achieve better results with custom designed speech processors for cochlear nucleus implants. Current research is concentrating on microstimulation of the ventral cochlear nucleus with penetrating micro-electrodes to help people who are completely deaf.
For patients with cochlear implants, Dr. Rubinstein strives to approximate as well as possible the effects that sound waves create on the hair cells of persons with normal hearing. In both cases, the sound is converted into an electrical signal to activate the auditory neurons via the hair cells or via the implant in the cochlea, the snail shell-shaped organ situated in the inner ear. From this point, the data is transmitted to the auditory brainstem. In the computer model, the response of 100 cells is simulated, each consisting of 24 nodes. The nodes propagate nerve impulses in much the same way as a repeater in a telephone line initiates and boosts the signal. Between two nodes, a row of nine cell segments attenuates the signal whenever it passes the internode sections. The entire model thus consists of 240 elements and each node provides 4000 sodium channels to produce the nerve impulse.
These figures explain the need for a supercomputer since the model only simulates the response to a 10-millisecond sound but in no less than 10.000 1-microsecond time steps. In order to study more complex speech patterns, Dr. Rubinstein requires sophisticated algorithms and higher performance hardware. To create a physiologically realistic model, a random number generator has been integrated into the model to prevent the sodium channels from reacting identically to the same stimulus over again. In a real auditory system, the sodium channels open and close stochastically, thus provoking a different response to the same stimulus. A similar effect is now incorporated in the simulation model. The results achieved both by animal testing and computational simulation will be to the benefit of future generation implants' design. Please, consult the Web sites of the National Institutes of Health and the University of Iowa for more information on cochlear implants.