Psychophysical tuning curves, computed tomography imaging and computational modeling
Behavioral thresholds and psychophysical tuning curves (a measure of channel interaction) obtained with focused electrical fields may be sensitive to the way each electrode interfaces to auditory neurons, known as the electrode-neuron interface. One aspect of the electrode-neuron interface is the distance of each electrode from the inner wall of the cochlea, near where the target spiral ganglion neurons reside. We can estimate this distance from a computerized tomography (CT) image, a type of x-ray image that can give fine detail about the cochlear implant array and how it is positioned within the cochlea. A recent comparison of electrode distances with threshold indicate that these measures are highly correlated, such that larger distances are associated with higher thresholds (DeVries et al, 2016; Bierer et al, 2014). Electrical impedances may also be higher for larger electrode-neuron distances (Bierer et al, 2014 [Obj. Meas]), suggesting that current flow is impeded for these electrodes. Another possible factor affecting the electrode-neuron interface is the viability of the spiral ganglion neurons, which can degrade over time depending on the progression of the hearing loss. We are developing a computational model that can help us evaluate the relative roles of neuron health, electrode position, cochlear impedances, and other factors in individual cochlear implant listeners. We are also investigating the use of electrically-evoked compound action potentials (DeVries et al, 2016). This type of evoked potential, which can be recorded directly from the implant without the need for specialized equipment, may be a quicker way to identify electrodes with poor neural interfaces compared to behavioral thresholds or, other behavioral measures.
Gap detection and rate discrimination
As mentioned previously, spectral resolution can be degraded for a subset of cochlear implant channels in some implant listeners. We’d also like to know if temporal resolution is compromised in these channels, as the proper processing of both spectral and temporal information is important for speech perception. One measure of temporal resolution is the minimum gap (in milliseconds) between two pulse trains that a listener can detect. Another measure is how well a listener can discriminate between two stimuli having different pulse rates (# of pulses per second): the more similar the pulse rates, the better the listener’s temporal acuity. We hypothesize that more severe degeneration of auditory neurons will result in poorer temporal resolution.
Evaluating the Consonant-Nucleus-Consonant (CNC) speech test
The Consonant-Nucleus-Consonant (CNC) word lists are commonly used in the clinic and research labs to determine how well a cochlear implant patient can understand speech. We are currently evaluating a way to improve this speech material by balancing the ten CNC word lists on the basis of lexical frequency, lexical density and phonetic content (Peterson and Lehiste, 1962). We are also examining an alternative way to score listener performance on the CNC word test, to better inform audiologists about speech sounds a patient might be struggling with and, subsequently, to optimize the programming of the implant speech processor.
Simulations of cochlear implant listening
A vocoder is a way to encode an acoustic sound such that a normal-hearing listener might hear it the same way that a cochlear implant listener might hear it. In our laboratory, we are assessing various manipulations of the vocoder algorithm to simulate the different patterns of neural survival that a cochlear implant listener may have. We are altering the vocoder in three ways: 1) setting a subset of noiseband outputs to zero amplitude, as way to simulate non-viable neurons; 2) forcing a subset of channels to share the same filter outputs, as a way to simulate a high degree of channel interaction; 3) presenting unmodulated noise to a subset of channels, as a way to simulate reduced-viable neurons. We will analyze the specific error patterns that a normal-hearing subject makes when listening to the vocoded speech stimuli, with the goal of relating the spectral content in the stimuli to the particular neural deficit being simulated.
Fitting of cochlear implants
Based on the results of the psychophysical tuning curve and threshold perceptual tests, evoked potential measures, CT imaging, computer modeling, and vocoder experiments, we will develop patient-specific methods for optimizing cochlear implant fitting. These methods will include the application of current focusing and steering in an attempt to target healthy populations of auditory neurons and improve the transmission of spectral and temporal information.