Reliable identification of click-evoked otoacoustic emissions using signal-processing techniques

As neonatal hearing screening programmes based on click-evolved otoacoustic emissions (CEOAEs) become more common, there is a need for automatic scoring methods to classify responses systematically. Such methods must guard rigorously against errors that would pass a baby who would otherwise fail and be referred for further diagnostic testing. Preliminary analysis of a wide range of CEOAE waveform characteristics from adult and neonatal ears suggested that the most reliable measure on which to base an automatic method was the cross-correlation coefficient between replicate non-linear components. Further analysis of recordings from 2049 ears of 1102 neonates in neonatal intensive care units indicated that a cross-correlation coefficient greater than 0.5 was an appropriate criterion to distinguish a pass from a fail, with the following two provisos. The waveforms should be high-pass filtered to remove random low-frequency components before calculating the cross-correlation coefficient. The cross-correlation coefficient should relate to the time period from 6-16 ms after click onset. Using this method, 90% of recordings were automatically classified in agreement with a proven method of visual scoring. Only one ear without evidence of otoacoustic emission was passed by the automatic scoring method.