Influence of signal processing on OAE-based newborn hearing screening

Click-evoked otoacoustic emissions (CEOAEs) are currently used in neonatal hearing screening programs for the early identification of hearing impairment present at birth. CEOAEs have been shown to be a rapid, cost-effective test with a high sensitivity but a low specificity: it fails a relatively large number of babies having a normal hearing sensitivity, CEOAE scoring is greatly influenced by the decision rules separating passes from fails, the acquisition procedure and the analysis software. On the other hand, no consensus exists regarding the numerical criteria to be used when scoring CEOAEs. Both simulated signals and CEOAEs from more than 3,400 newborns were here used to examine the effects that different data processing options, such as time-windowing, digital band-pass filtering, type of synchronous averaging, and frequency decomposition, have on the specificity of the test. The specificity was found to vary from less than 40% to more than 95%, depending on the decision rules and on the type of signal processing used. In most cases, the procedures used by defaults by the most widespread equipment for CEOAE recording were relatively far away from the optimal solution and that there is still a wide possibility to increase the accuracy of the test