Reduction of listening effort with speech enhancement strategies in cochlear implant users

A cochlear implant (CI) is a prosthetic device for the inner ear, able to
directly stimulate the auditory nerve. With a CI it is possible to bypass
damaged inner hair cells and provide audible sensations to profoundly deaf
implant recipients. Hearing in quiet with a CI is generally sufficient with
word recognition scores approaching 100% for the best performing CI users.
However, listening in noise degrades this speech understanding substantially,
and more so when compared to normal-hearing listeners. Much of the present
research focuses on improving speech understanding (SU) in noise by means of
noise reduction algorithms in the speech processing chain of the device.
Meanwhile, listening effort (LE) has been largely neglected, but is receiving
increasing attention in recent years. Due to the advances in technology, CI
users are increasingly able to understand the desired sounds in adverse
listening conditions. However, listening in such challenging conditions leads
to listening fatigue in CI users. Because fatigue decreases speech recognition
performance, it is critical to minimize LE in CI users. This proposal will
focus on measuring LE and the effectiveness of different SESs in reducing LE.

Measuring LE alongside SU is a powerful combination. Generally,
speech-enhancement strategies (SESs), such as noise reduction algorithms, are
tested for effectiveness only by assessing SU scores. An often overlooked fact
is that if the test subject consciously or unconsciously puts less effort in
the task when the SES is switched on, then SU may not improve at all, hence
underestimating the true benefits of the noise reduction strategy.

One method to measure LE is pupillometry, which is rapidly gaining popularity
in recent years. Pupillometry assesses the pupil diameter, which is a proxy for
cognitive load, and hence a useful outcome measure to assess LE.

There are 2 main limbs to this study:

1. Create a benchmark, in which we will assess the three most commonly used
Dutch speech-in-noise tests, namely the LIST, Matrix, DIN and Plomp for their
associated LE in noise. This is necessary as pupillometry provides a measure of
cognitive load; the LIST and Plomp sentences are meaningful sentences where it
has been hypothesized that more mental load is necessary than in meaningless
sentences as applied in the Matrix and DIN tests. In our lab we use the Matrix
and LIST tests, and we wish to compare the performance of these tests against
the DIN and Plomp sentences. For the Plomp sentences, LE has been firmly
established [2, 3].

2. We will test a number of promising SESs using pupillometry, including
2a. SoftVoice: diminishes system noise in the CI and allows for better SU
especially at low sound levels.
2b. Dereverberation algorithm: diminishes reverberations which enter the CI.
And creates therefore a better speech understanding in an environment with lots
of reverberations, e.g. a swimming pool or a restaurant.
2c. Bimodal hearing: The addition of a hearing aid to the CI has been tested
previously in a successful study. However, at the time of that study we were
not able to perform pupillometry. While our subjects were overwhelmingly
positive about the bimodal hearing solution (added a hearing aid), the overall
improvement in speech understanding was marginal (yet significant). We
hypothesize that the main benefit of bimodal hearing is to be sought in LE, and
not in SU per se.

This will be an observational cross-over trial. The total duration of the study
is estimated to be 4 years. Each subject will typically participate in one
limb, that is, for 4 sessions, but they are free to participate in more limbs.
Each session lasts approximately 2.5 hours. Optionally, an SES may be fitted on
their home-use CI. In that case subjects will be asked to use the SES for
anywhere between a few days to a few months at home. Hence, subjects may be
enrolled in this ongoing study for a few months, during which they will have to
come in to undergo speech testing, typically for 4 sessions. They will
participate in more sessions when they participate in multiple sub-studies, but
nor more than 12 sessions in total to limit the burden, except when they insist
to complete all substudies. The setting is in the LUMC. It is a cross-over
trial as every subject will be their own control when an SES is tested: SES ON
(test) versus OFF (control) (limb 2), or when speech tests are compared (limb
1).

Minimal burden and negligible risk