Understanding Aided Speech Perception in Noise

Description

STUDY DESIGN:

This is a between-group and within-subject, repeated-measures design. A detailed protocol, visit schedule, intervention methodology, and battery of outcome measures will be used to assess audition, spatial sensitivity, and pre- and post-intervention of behavioral and neurophysiological indices of spatial segregation. The study includes two control groups (young normal-hearing [YNH] and older normal-hearing [ONH]) and the test group (older hearing-impaired [OHI]) with three aided interventions: omnidirectional processing, fixed directional processing, and variable directional processing. There are two study arms: Arm 1 tests spatial segregation in the presence of a fixed masker in front of the listener; Arm 2 tests spatial segregation in the presence of a diffuse masker presented from three (narrow) or seven (wide) loudspeakers surrounding the frontal spatial hemisphere.

DESCRIPTION OF THE PROTOCOL:

All listeners receive an intake protocol consisting of standard clinical and laboratory measures of auditory function (roughly two-hour visit). The baseline measures include audiogram, loudness discomfort levels (LDLs), middle ear status (tympanometry), acoustic reflex thresholds, and words in noise. Pre-treatment measures of spatial sensitivity are measured in a single two-hour visit. Measures include the minimum audible angle (MAA), speech localization error, and spatial release from masking. The Pre-treatment measures of the spatial segregation boundary are conducted in a single two-hour visit. Post-treatment measures (OHI group only) are counter balanced across three additional two-hour visits. Pre- and post-treatment measures of the spatial segregation boundary include measures of a fixed spatial separation between target and masker speech, and a roving (moving) target condition at multiple spatial separations spanning 0° to 90° in 15° steps.

DELIVERY OF INTERVENTION:

Hearing aid fitting: The standard-of-care intervention for ARHL is amplification via hearing aids and this is the intervention to be used in this study. The fitting procedures and any necessary troubleshooting or adjustments will follow the Auditory Neurosciences & Technology Laboratory device Standard Operating Procedures (dSOP) manual for hearing aid fitting and counseling. This manual is based upon guidelines of the American Speech-Language-Hearing Association and the American Academy of Audiology. Devices are fit wirelessly using a clinical-style graphical user interface for device programming in conjunction with real-ear verification. The audiologist (Dr. Secor) will fit the OHI subjects with the hearing instruments prior to the experiments. To avoid the possibility that independent amplitude compression at the two ears can reduce ILD cues, aids will be fit with a linear prescription. Because the sound levels in the experiments are in a restricted range near 65 dB the investigators will use NAL2 prescription for 65 dB input levels as verified using probe-microphone measures (Verifit 2, Audioscan).

Hearing aids: In order to better generalize results and control for a wide array of potentially hidden digital processing effects by device manufacturers, the investigators will use open-source, research-grade devices made possible through the NIH-funded open Master Hearing Aid consortium (openMHA). The open-source hardware design includes behind-the-ear (BTE), receiver-in-the-canal (RIC) instruments (Sonion) wired to a multi-channel audio board (Cape4all; 16 kHz sampling) attached to a portable main board (Beaglebone Black Wireless). The lightweight hardware is worn around the user’s neck. Beamforming (fixed or variable) is accomplished in openMHA using a binaural minimum variance distortionless response (MVDR) algorithm that attempts to minimize any diffuse or spatially separated noise contributions from a target speech source.

Hearing aid conditions: In a within-subject design, the OHI group will be tested (per study arm) in four device configurations: (1) unaided, (2) aided with omnidirectional microphones, (3) aided with fixed directional processing, and (4) aided with variable directional processing. The fixed binaural beamformer has a relatively narrow spatial filter at the front, whereas the variable binaural beamformer is steered to the speech direction in real time. For the variable condition, real time speech localization is based on a priori knowledge of where the speech is presented from instead of employing an adaptive speech locator as used in real devices. This condition, therefore, represents the best achievable spatial filtering given perfect source localization and zero lag time between source localization and beam steering.

SAMPLE SIZE AND DATA ANALYSIS METHODS:

Power analyses. Based on preliminary data and reports with similar analyses in the literature, significant effect sizes are expected to have a minimum of ηp2 = 0.19 for repeated measures analysis of variance (rmANOVA) testing. In Arm 1, computing sample size for the smallest possible effects (ηp2 = 0.19), alpha (0.05), power (0.95), with moderate-to-strong correlation among repeated measures (0.75) indicates that 38 listeners per group are needed for each experiment (≥3 repeated measures; G*Power v3.1.9.2). In Arm 2, the investigators will also use multiple regression to measure the relationship between spatial acuity and hearing aid benefit. For a small-to-medium effect size (f2 = 0.3), 55 subjects are needed.

Analytic plan. Descriptive and graphical statistics will be used to summarize the data on all participants and appropriate transformations, or non-parametric methods will be applied, as necessary. The investigators will use rmANOVAs to analyze the results from the behavioral data and cortical synchronization results at the group level. Pearson’s correlation coefficients will be Z-transformed (Fisher) before statistical analyses. To protect against Type I error, the principal analyses are designed to answer clearly stated major research hypotheses and are based on well-specified dependent variables. All tests will be two-tailed at alpha=0.05, with Holm-Bonferroni adjustment for multiple comparisons. Where appropriate, rmANOVA’s will be used to measure group differences in our intake psychoacoustic measures, and if the investigators observe statistical group differences, they will be used as covariates.