Detection of sensorineural hearing impairment by classifying latency and amplitude parameters of DPOAE`s

Abstract

Both amplitude and latencies of Distortion Product Otoacoustic Emissions(DPOAE) at 2F1-F2 are measured in response to pure tones at F1 and F2 over the range 1-9 KHz and 1-6 KHz of F2 respectively, from 86 normally hearing and 22 hearing impairedhuman ears.The normality ranges are determined for amplitude (DPgram) and latency values ofDPOAEs, for future comparison with the pathologic group data results. DPgram charts ofhearing impaired subjects are investigated to demonstrate any correlation between DPOAEsand pure tone audiograms. Moreover, latency values are used to obtain information aboutthe generation site of2F1-F2 distortion products on the basilar membrane of cochlea. Whenhearing loss was worst than 50 dB HL, DPOAE was rarely detected.K-means and artificial neural network (ANN) classifiers are implemented to find thebest feature of DPOAEs in classifYing normal and pathological ears. For this purpose,amplitude and latency parameters ofDPOAEs are either used separately or together as inputfeatures to the classifiers, and the success rates of classifiers for different feature selectionsare determined. Also, for the ANN classifier, different network configurations are used toreach to the best classification performance. It has been shown that, if both amplitude andlatency parameters are considered simultaneously, the rate of success is higher.