Time-efficient behavioural estimates of cochlear compression

Currently, the main tool used to diagnose and fit hearing aids for people with sensorineural hearing loss is based on the individual’s frequency dependent sensitivity, the audiogram. While useful, the audiogram does not reflect any suprathreshold performance. Thus, certain hearing-aid settings may be optimal for one listener and suboptimal for another, even if the audiograms of the two listeners are similar.

It has been suggested that knowledge of the nonlinear characteristics of an impaired cochlea could improve rehabilitation through individualized hearing-aid fitting. However, behavioural experiments that have been considered to enable such an extended characterization are too time-inefficient to be relevant for clinical practice.

In this thesis, a new threshold tracking method, called the “Grid” method, is developed. The main difference between the Grid and the traditional methods is the way of sampling a threshold function. While traditional approaches vary just one parameter during a single experimental run, the Grid method enables varying more than one parameter.

This increases the proportion of the experimental time spent in the vicinity of the threshold function. The first part of the thesis establishes the basis for the development of the Grid method. Two approaches to improve the time-efficiency of forward-masking experiments, and in particular, so-called temporal masking curves, were investigated.

In both approaches, the single-interval up-down threshold-tracking method was used. In the first approach, the masker-level thresholds were found for a set of masker-signal gaps. In the second approach, the masker-signal gap thresholds were found for a set of masker levels. The results suggested that it was possible to derive estimates of outer hair cell loss, based on the estimated thresholds.

However, practical limitations limited the accuracy of the loss estimates in both methods. Moreover, the methods used in the two approaches showed complementary advantages, which could not be combined in a simple way. Next, the study defined the assumptions, hypotheses and technical details of the Grid method and compared the thresholds obtained with this method to those obtained with standard methods.

The results showed strong correlation between the corresponding thresholds. While a detection task was considered in this part, the next part applied the Grid method to a discrimination task to explore the compression characteristics at low frequencies. The obtained results were in good agreement with the corresponding results in the literature.

Finally, numerical simulations were performed to assess the accuracy of the Grid method in terms of bias and variance. Overall, the results demonstrated that theGrid method is capable of delivering a similar accuracy as current behavioural methods in a fraction of the time. While the exact time-efficiency advantage can vary based on paradigm details, a reduction in testing time by an order of magnitude can be achieved.

These results may be very useful in both laboratory as well as clinical settings, where time-efficiency is an important factor.