Automatic screening and detection of threshold fine structure

Abstract

Audiograms measured with a high frequency resolution often show quasi-periodic ripples of up to 15dB in normal-hearing listeners. This fine structure of the threshold in quiet is commonly associated with the active processes in the cochlea. Therefore its absence is discussed in the literature as an indicator of cochlear vulnerability. In order to enable a quick detection and an objective quantification of threshold fine structure, two instruments are introduced and evaluated in this article: (1) a high-resolution tracking method for measuring fine structure (‘FINESS’), and (2) an automatic fine-structure detector (‘FINESS-detector’). The method is tested on 22 subjects for its reliability, its accuracy, and drifts with frequency by analysing test/retest experiments and by comparing the measured thresholds to results from a reference procedure. The results indicate that FINESS and the FINESS-detector are suitable techniques for the measurement and detection of threshold fine structure that may help to investigate further into whether fine structure is a sensitive tool for the detection of an early hearing loss.

• Threshold fine structure
• Audiological methods

Abbreviations
AFC	=	Alternative forced choice
DPOAE	=	Distortion product otoacoustic emission
EOAE	=	Evoked otoacoustic emission
OAE	=	Otoacoustic emission
SOAE	=	Spontaneous otoacoustic emission

Abbreviations
AFC	=	Alternative forced choice
DPOAE	=	Distortion product otoacoustic emission
EOAE	=	Evoked otoacoustic emission
OAE	=	Otoacoustic emission
SOAE	=	Spontaneous otoacoustic emission

Sumario

Los audiogramas medidos con resolución de alta frecuencia a menudo muestran ondas cuasi-periódicas de hasta 15 dB en individuos normo-oyentes. Esta estructura fina del umbral en silencio se asocia comúnmente con los procesos activos en la cóclea. Por lo tanto, su ausencia se discute en la literatura como un indicador de vulnerabilidad coclear. Para facilitar una detección rápida y una cuantificación objetiva de la estructura fina del umbral, se han introducido y evaluado dos instrumentos en este artículo: (1) un método de rastreo de alta resolución para medir dicha estructura fina (“FINESS”), y (2) un detector automático de estructura fina (“Detector FINESS”). El método se evaluó en 22 sujetos con respecto a su confiabilidad, su exactitud y a las variaciones de frecuencia, por medio del análisis de experimentos test/retest y de la comparación de los umbrales medidos con los resultados de un procedimiento de referencia. Los resultados indican que el FINESS y el detector FINESS son técnicas apropiadas para la medición y la detección de la estructura fina del umbral, que puede ayudar a investigar si la estructura fina es una herramienta sensible para la detección temprana de una pérdida auditiva.

Notes

1. Subjects with thresholds in quiet of up to 15 dB HL according to their audiogram are generally considered to have normal hearing. At 8 kHz three subjects (HJ, HS, NM) showed thresholds of 20 dB HL, and one subject (HR) of 25 dB HL, in one ear. This may be interpreted as a mild hearing loss at this high frequency, although some studies even consider a threshold of 25 dB HL as still normal. In addition, it is possible that some of the subjects may already have an initial hearing loss in the relevant frequency range of 1–3 kHz which is undetectable by standard audiometry. The aim of the present study is to develop a reliable method for measuring fine structure in order to help to possibly establish a link between such early hearing losses and the amount of threshold fine structure in future studies. Thus, for evaluating the method it is not crucial that all subjects have normal hearing, as long as some of them show fine structure and some do not. In order to enhance the probability of finding threshold fine structure, only subjects with no (or only a mild) hearing loss in their pure-tone audiograms were chosen for the present study.

2. Correlation coefficients in this study were averaged by applying Fisher's transformation Z=artanh(r), then calculating the arithmetic mean of the transformed coefficients and transforming it back. This is done because coefficients of correlation are not values on a scale of equal metric units. Differences between large r's are actually much greater than those between small r's. Therefore the sampling distribution of the correlation coefficient is skewed so that the arithmetic mean of correlations would underestimate their central tendency (Guilford & Fruchter, 1973; Silver & Dunlap, 1987).

3. Here an adaptation of the ‘Audioscan’ method is referred to in which the frequencies are swept slow enough to detect fine structure.