Multimodal artificial intelligence (AI) could support audiology reporting, but an end-to-end score can obscure whether errors arise from image transcription, specialty-rule execution, or communication. We evaluated these as separate modules in 155 de-identified outpatient records representing 151 patients....
✦ The floor
Discussion
Signed responses from readers of the wire.
Audiologist-guided AI shows early promise for automating audiometry reporting, but error analysis across pipeline stages means it is not yet ready to replace clinician review; treat as a research prototype only.
AI-assisted interpretation and reporting of routine hearing assessments could reduce administrative burden and improve consistency in audiology clinics, if accuracy and safety are validated at scale.
- 01A multimodal AI system was developed to interpret pure-tone audiometry and tympanometry results and auto-generate reports.
- 02Audiologists guided the AI design, and errors were categorized across transcription, rule execution, and communication stages.
- 03Identifying error sources is essential before clinical deployment of AI reporting tools.
- 04The approach targets two of the most commonly performed tests in audiology clinics.
- 05Published in Journal of Medical Systems (doi: 10.1007/s10916-026-02463-5).
A multimodal AI system guided by audiologists can interpret and report pure-tone audiometry and tympanometry results.
studypartially supportedErrors in AI audiometric reporting can be traced to distinct pipeline stages: transcription, rule execution, and communication.
studysupported- PMID
- 42821199
- DOI
- 10.1007/s10916-026-02463-5.
- Journal
- Journal of Medical Systems
- Publication type
- research_article
- Evidence level
- 4
- Population
- Audiometric test results interpreted by an AI system under audiologist guidance
- Intervention
- Multimodal AI system for pure-tone audiometry and tympanometry interpretation and reporting
Primary outcomes
Accuracy of AI-generated audiometric reports; Error source categorization across transcription, rule execution, and communication stages