Evaluation of a Fast Method to Measure High-Frequency Audiometry Based on Bayesian Learning

Chiara Casolani, Ali Borhan-Azad, Rikke Skovhøj Sørensen, Josef Schlittenlacher, Bastian Epp*

*Corresponding author for this work

Research output: Contribution to journalJournal articleResearchpeer-review

11 Downloads (Pure)

Abstract

This study aimed to assess the validity of a high-frequency audiometry tool based on Bayesian learning to provide a reliable, repeatable, automatic, and fast test to clinics. The study involved 85 people (138 ears) who had their high-frequency thresholds measured with three tests: standard audiometry (SA), alternative forced choice (AFC)-based algorithm, and Bayesian active (BA) learning-based algorithm. The results showed median differences within ±5 dB up to 10 kHz when comparing the BA with the other two tests, and median differences within ±10 dB at higher frequencies. The variability increased from lower to higher frequencies. The BA showed lower thresholds compared to the SA at the majority of the frequencies. The results of the different tests were consistent across groups (age, hearing loss, and tinnitus). The data for the BA showed high test–retest reliability (>90%). The time required for the BA was shorter than for the AFC (4 min vs. 13 min). The data suggest that the BA test for high-frequency audiometry could be a good candidate for clinical screening. It would add reliable and significant information without adding too much time to the visit.

Original languageEnglish
JournalTrends in Hearing
Volume28
Pages (from-to)1-12
ISSN2331-2165
DOIs
Publication statusPublished - 2024

Keywords

  • Audiometry
  • Bayesian learning
  • High-frequency
  • Synaptopathy

Fingerprint

Dive into the research topics of 'Evaluation of a Fast Method to Measure High-Frequency Audiometry Based on Bayesian Learning'. Together they form a unique fingerprint.

Cite this