gms | German Medical Science

The pure-tone audiogram is the main clinical diagnostic used for assessing hearing loss and provides the basis for prescribing hearing-aid amplification to restore audibility. However, the audiogram does not necessarily reflect the hearing-in-noise deficits that can remain when audibility has been restored, which can manifest themselves, e.g., in poor speech understanding in adverse conditions. A clinical measure of such deficits would thus be highly useful for prescribing advanced hearing-aid features designed to counteract hearing-in-noise deficits, such as powerful deep neural network (DNN) based noise reduction algorithms.

To enable assessment of hearing-in-noise deficits in the clinic, we have conducted multiple research studies optimizing spectro-temporal modulation (STM) detection tests for hearing-impaired listeners such that the measured STM performance predicts the listeners’ speech-in-noise performance [1], [2]. These studies resulted in the development of a clinical test paradigm called the Audible Contrast Threshold (ACTTM) test, a novel quick-and-simple clinical STM detection test with built-in audibility compensation [3].

We evaluated ACT in a large-scale clinical study with 100+ participants in terms of test-retest reliability, testing time, and predictive power with regard to aided speech-in-noise performance. Speech reception thresholds (SRTs) were measured with participants using Oticon More 1 hearing aids in a challenging setting with spatially distributed speech interferers. Four different hearing-aid settings were tested: amplification only, mild directionality and noise reduction (DIR+NR), medium DIR+NR, and strong DIR+NR, using Oticon’s DNN-based help-in-noise feature.

The ACT test was found to be quick (<2 minutes) and reliable. On the group level, SRTs were highest for the amplification-only setting and decreased with increasing levels of DIR+NR processing. The individual SRTs collected with amplification only were strongly correlated with ACT and – to a lesser extent – with the 4-frequency pure-tone average (PTA4). The predictive power of ACT and PTA4 was found to be complementary, as they both contributed significantly to predicting the amplification-only SRTs in a two-predictor linear regression model. The individual SRT benefit induced by different levels of DIR+NR varied substantially across participants, with poor performers benefiting most from strong DIR+NR. A dedicated analysis showed that the variability in SRT benefit was correlated with the closedness of acoustic coupling and with ACT.

Overall, the results suggest that ACT is a clinically viable hearing-in-noise test that can be used to prescribe advanced hearing-aid features, especially DIR+NR features that improve the speech-to-noise ratio. Special attention should be paid to the effects of acoustic coupling on the real-life efficacy of DIR+NR processing.