Objective Assessment System for Hearing Prediction Based on Stimulus-Frequency Otoacoustic Emissions.
Qin Gong, Yin Liu, Runyi Xu and 3 others
PMID 34817273WHAT IT FOUND
Machine-learning models using stimulus-frequency otoacoustic emissions estimated hearing thresholds within 10 dB in 62.05% to 83.71% of ears and identified hearing status in 88.38% to 95.93% of ears, but thresholds of 60 dB HL or worse were not predictable.
Key findings
01The routine module predicted hearing thresholds with mean absolute errors from 7.06 dB at 1 kHz to 11.61 dB at 8 kHz, and 62.05% to 83.71% of ears were predicted within 10 dB.
02The fast module took 3.2 to 4.0 minutes per frequency and did not show statistically significant differences in mean absolute errors compared with the routine module.
03The hearing screening module correctly identified hearing status in 90.82% to 95.93% of ears at 0.5 to 4 kHz and 88.38% at 8 kHz, with false negative rates of 2.87% to 7.02%.
STILL TO COME
How it was doneWhat they foundWhat it means for SLPs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The system was built from custom hardware and software for research use, and the paper states that clinical instruments for recording SFOAEs were not available. Threshold prediction was worse in ears with sensorineural hearing loss than in normal-hearing ears, and worse at 0.5 and 8 kHz than at 1 to 4 kHz. The fast module could not use its shorter stopping rule in many ears with sensorineural hearing loss, because only 34.06% to 53.28% met the signal-to-noise criterion. Hearing thresholds of 60 dB HL or worse could not be predicted, because SFOAEs were often absent when hearing loss was severe. Testing was done in a sound-attenuating chamber by a research assistant, not in a clinical care setting. The models were frequency-specific, so an ear could be classified as normal at some frequencies and hearing loss at others. The authors note that coupler calibration and high stimulus levels may affect accuracy.
Declared interests
The authors declared no potential conflicts of interest. The work was supported by the National Natural Science Foundation of China (grant number 61871252) and the Foundation of Jiangsu Province Science and Technology (grant number BE2020635).
The easy way to misread this
Do not read this as a hearing test ready for clinical use. The models were trained and tested in a research setting on ears selected for normal middle ear function, and the system could not predict thresholds of 60 dB HL or worse.