Transiently evoked otoacoustic emissions (TEOAEs) are routinely used in the hearing assessment of the auditory periphery. The major contribution of TEOAEs is the early detection of hearing losses in neonates, children, and adults. The evaluation of TEOAE responses by specific signal decomposition techniques offers numerous advantages for current and future research. One methodology, based on recurrence quantification analysis (RQA), can identify adult subjects presenting sensorineural hearing impairments. In two previous papers, the RQA-based approach was succesfully applied in identifying and classifying cases presenting noise and age related hearing losses. The current work investigates further two aspects of the previously proposed RQA-based analysis for hearing loss detection: (i) the reliability of a Training set built from different numbers of ears with normal hearing, and (ii) the threshold set of values of the key hearing loss detecting parameter RAD2D. Results: The Training set built from 158 healthy ears was found to be quite reliable and a similar but slightly minor performance was observed for the training set of 118 normal subjects, used in the past; the proposed ROC-curve method, optimizing the values of RAD2D, shows improved sensibility and specificity in one class discrimination. Conclusions: A complete and simplified procedure, based on the combined use of the traditional TEOAE reproducibility value and on values from the RQA-based RAD2D parameter, is proposed as an improved automatic classifier, in terms of sensitivity and specificity, for different types of hearing losses.
Detection of hearing losses (HL) via transient-evoked otoacoustic emissions: towards an automatic classification
Giovanna Zimatore
;
2022-01-01
Abstract
Transiently evoked otoacoustic emissions (TEOAEs) are routinely used in the hearing assessment of the auditory periphery. The major contribution of TEOAEs is the early detection of hearing losses in neonates, children, and adults. The evaluation of TEOAE responses by specific signal decomposition techniques offers numerous advantages for current and future research. One methodology, based on recurrence quantification analysis (RQA), can identify adult subjects presenting sensorineural hearing impairments. In two previous papers, the RQA-based approach was succesfully applied in identifying and classifying cases presenting noise and age related hearing losses. The current work investigates further two aspects of the previously proposed RQA-based analysis for hearing loss detection: (i) the reliability of a Training set built from different numbers of ears with normal hearing, and (ii) the threshold set of values of the key hearing loss detecting parameter RAD2D. Results: The Training set built from 158 healthy ears was found to be quite reliable and a similar but slightly minor performance was observed for the training set of 118 normal subjects, used in the past; the proposed ROC-curve method, optimizing the values of RAD2D, shows improved sensibility and specificity in one class discrimination. Conclusions: A complete and simplified procedure, based on the combined use of the traditional TEOAE reproducibility value and on values from the RQA-based RAD2D parameter, is proposed as an improved automatic classifier, in terms of sensitivity and specificity, for different types of hearing losses.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
