Estimating Hearing Thresholds From Stimulus-Frequency Otoacoustic Emissions.

Area of Science:

• Audiology
• Otoacoustic Emissions
• Hearing Science

Background:

• Estimating pure-tone thresholds from objective measures like stimulus-frequency otoacoustic emissions (SFOAEs) is clinically valuable.
• While SFOAEs can indicate hearing status, their potential for predicting specific audiometric thresholds remains underexplored.

Purpose of the Study:

• To investigate the efficacy of SFOAEs in predicting hearing thresholds across octave frequencies (0.5–8 kHz).
• To compare a novel back propagation (BP) neural network predictor with a linear regression model for threshold prediction.
• To evaluate the performance of a BP network classifier in identifying hearing status.

Main Methods:

• Collected SFOAE input/output functions and pure-tone thresholds from 230 normal-hearing ears and 737 ears with sensorineural hearing loss.
• Developed and applied a linear regression model (Method 1) and a BP neural network predictor incorporating principal component analysis (Method 2) to estimate thresholds.
• Utilized a BP network classifier to determine hearing status.

Main Results:

• Both prediction methods successfully estimated hearing thresholds from 0.5 to 8 kHz.
• Method 2 (BP network predictor) significantly outperformed Method 1 (linear regression).
• BP network classifiers demonstrated excellent accuracy in identifying hearing loss across all tested frequencies, with optimal prediction for thresholds between 0.5 and 4 kHz.

Conclusions:

• SFOAEs accurately identify hearing status and can predict hearing thresholds, particularly between 0.5 and 4 kHz.
• The BP neural network predictor shows promise as a quantitative tool for estimating hearing thresholds.
• SFOAEs offer a valuable, non-invasive method for audiological assessment and hearing threshold prediction.