Multifaceted evaluation of a binaural coclear-implant sound-processing strategy inspired by the medial olivocochlear reflex

Laburpena

Cochlear implants (CIs) can enable useful hearing to deaf persons via direct electrical stimulation of the auditory nerve. Despite the progress achieved in CI design and performance, CI users still struggle understanding speech in noise or localizing sound sources with modern, bilateral CIs (BiCIs). The MOC strategy is a binaural CI sound coding strategy inspired by the dynamic control of basilar membrane (BM) compression provided in natural hearing by the contralateral medial olivocochlear reflex (MOCR). In contrast to the standard clinical approach (STD), which involves using two independently functioning audio processors with fixed acoustic-to-electric compression, the MOC strategy dynamically couples the amount of compression applied in each ear. This can result in better speech-in-noise recognition [Lopez-Poveda et al., 2016, Ear Hear 37(3):e138-148]. Though promising, the original MOC strategy had potential drawbacks and its parameters disregarded important aspects of the natural MOCR. The aim of this thesis was to experimentally investigate the potential additional benefits of more realistic implementations of the MOC strategy for speech-in-noise recognition, sound source localization, and listening effort. The thesis comprises four studies. The first study focused on speech-in-noise recognition. Speech reception thresholds (SRTs) for sentences presented in competition with steady-state noise were measured in unilateral and bilateral listening modes, and for multiple spatial configurations of the speech and noise sources. Speech reception thresholds were compared for stimuli processed through a STD strategy; the original MOC strategy with fast control of compression and greater inhibition at higher than at lower frequencies (MOC1); a MOC1 strategy with slower control of compression, thus closer to the time course of MOCR inhibition (MOC2); and a MOC2 strategy with greater inhibition at lower than at higher frequencies (MOC3), thus closer to the MOCR. We found that the more realistic MOC3 strategy overcame the shortcomings of the original MOC1 strategy and provided overall better speech-in-noise recognition. In addition, the MOC2 and MOC3 strategies provided a significant binaural advantage, which was not the case for the other strategies tested. The second study focused on sound source lateralization. Bilateral CI users were asked to localize noise tokens in a virtual horizontal plane for stimuli processed through the STD, MOC1, MOC2, and MOC3 strategies. Compared to the STD strategy, the MOC1 strategy slightly improved the localization of broadband noise bursts 200 ms in duration. The MOC2 and MOC3 strategies did not improve localization because stimuli were too short to fully activate and deactivate the contralateral control of compression but could theoretically provide similar improvements for longer stimuli as the MOC1 strategy did for shorter stimuli. The third study was aimed at investigating the potential benefits of combining MOC3 processing with a coding strategy (termed FS4) intended to preserve auditory temporal fine structure (TFS) cues in the four most apical frequency channels. Speech reception thresholds for sentences processed through the MOC3-FS4 and a standard FS4 strategy (STD-FS4) were compared in quiet, in steady-state and fluctuating noise, for various speech levels, in bilateral and unilateral listening modes, and for multiple spatial configurations of the speech and noise sources. Overall, SRTs were equal or better with the MOC3-FS4 than with STD-FS4 strategy. The fourth study was aimed at investigating if recognizing speech in noise was as effortful with the MOC strategies as it was with the more conventional STD strategies. Word recall scores and verbal response times in a word recognition test were used as proxies for listening effort and were measured in quiet, in steady-state noise at +5 dB signal-to-noise ratio (SNR) and at the individual SRT for sentences in noise. The results showed that BiCI users experienced approximately the same effort with all sound-processing strategies. Together, the findings show that the binaural MOC strategy, with realistic MOCR parameters, can improve sound-source localization and speech-in-noise recognition without increasing listening effort. In addition, they show that it is possible to combine MOC processing with state-of-the-art fine-structure audio coding for CIs, making the MOC strategy a promising approach to improve CI outcomes. Keywords: cochlear implants, olivocochlear efferents, dynamic-range compression, noise, speech intelligibility, sound localization, listening effort, audio coding.