FREQUENCY SPECIFIC AUDITORY BRAINSTEM RESPONSE IN …



A FUZZY EXPERT SYSTEM ANALYZING BROAD BAND AND FREQUENCY SPECIFIC - STIMULI RESPONSES

C.M. Koutsojannis1 and I.K. Hatzilygeroudis2

1. Department ofNursing, School of Health, Patras Highest Institute of Education & Technology

2. Department of Computer Engineering & Informatics, University of Patras

ABSTRACT

Latencies form the essence of most Auditory Brainstem Response (ABR) studies. Latency is best specified when a transient “click” stimulus is used. A click stimulus however is a broad-spectrum stimulus and thereby may stimulate the entire cochlear partition. Therefore it is difficult to separate frequency related influences on the ABR latencies. In the present study auditory brainstem responses using frequency-specific stimuli; were used in order to provide an approximation of the degree and configuration of cochlear hearing losses as well as the location of lesions higher in the auditory pathway. In this study a rule based fuzzy expert system was used to analyze the recordings of ABRs. Narrow and wide band stimuli were used to record responses that were obtained from 50 normal subjects, 20 subjects having various cochlear hearing losses for a limited frequency (1-2 KHz and 4-8 KHz), and 5 patients with Acoustic Neuroma. The latencies of waves I, III and V as well as the interpeak latencies I-III and III-V were analyzed and compared. The possibility to distinguish peripheral from central lesions through an intelligent multiparametric analysis of the aforementioned data is discussed on the basis of the tonotopical organization of the auditory system and different travelling time for the various frequencies.

Keywords: frequency-specific stimuli, evoked potentials, brainstem, expert system.

INTRODUCTION

The value of the Auditory Brainstem Response (ABR) in the detection and assessment of hearing and neurological disorders could be characterized in terms of sensitivity and specificity. ABR can be used in a variety of situations such as studies of the development of the auditory system, diagnosis of conductive versus sensorineural hearing loss, detection of pontine angle tumors and differential diagnosis of brainstem lesions. Several attempts to improve the sensitivity of ABR have been reported. These include: 1) the reduction of systematic sources of variance of the normative data depending on various factors such as age [Hecox & Galambos, 1974], sex [Beagley & Sheldrake, 1978], head size [Nikiforidis et al, 1993], etc. 2) the placement of the differential electrode pair in order to optimize ABR clarity, taking into account the spatial distribution of the ABR and noise [Starr & Squires, 1982]. 3) the use of appropriate stimuli in order to elicit the clearest ABRs [Chiappa, 1985, Starr & Don, 1988]. In all instances the authors used the click stimulus, a logical choice, since gross ABR development depends on neural synchrony. Click is a smoothed acoustic waveform which has a wide-band spectrum, i.e., contains energy up to about 10 KHz centered to the low frequencies. Its rapid onset and wide bandwidth excite the cochlear in a broad and synchronized way giving a strong volley of the nerve impulses to more rostral structures [Eggermont & Don, 1982, Eggermont, 1979]. On the other hand, if the goal is to distinguish between different pathological states or identify the location and extension of the lesion, the analysis of ABR using broad spectrum stimulus, presents serious limitations. For instance in cases with severe hearing loss in the 2 KHz region, absence of the ABR has no diagnostic significance. Also many different pathologic conditions present similar patterns of ABR as in the case of a small acoustic tumor and Multiple Sclerosis (MS) which may present with normal wave I and prolonged wave III and wave V latency [Musiek & Gollegly, 1985]. Finally, a high frequency hearing loss may be the result not only of peripheral sensory-neural lesion but also of lesions in the central auditory pathways [Dublin, 1976, Beattie et all 1996]. The aforementioned limitations of the ABR using broad spectrum stimulus (click), are due to the fact that the information expressed by the peak amplitudes and latencies cannot be easily related to the neuroanatomical substrate. The working hypothesis of most studies that the generators of the first five ABR peaks are: the cochlear nerve, the cochlear nucleus, the superior olive, the lateral lemniscus and the inferior colliculus, may be not valid [Hashimoto, 1982, Scherg & Von Grammon, 1985]. Indeed, there is experimental evidence supporting the view that each of these peaks has several anatomical contributions [Moller et al, 1981, Moller et al, 1988]. Anathanarian et al [Anathanarayan & Durant, 1991] reached this conclusion by ABR recordings using vertical and horizontal derivations of responses to clicks presented at a variety of stimulus levels and rates [Davis, 1984]. The neuroanatomical substrate explaining these experimental results is considered to be the tonotopical organization of the auditory system. In order to overcome the limitations regarding the localization and distinction of lesions along the auditory neural pathway, an alternative approach, introducing the use of frequency-specific stimuli in addition to that of the clicks, can be proposed. For better frequency definition of stimuli, different tone pulses [Suzuki & Horiuchi, 1981, Fausti et al, 1992, Buckard, 1991, Eggermont & Don, 1980], various selective masking techniques [Hoke et al, 1991], and deconvolution techniques on the click-evoked ABR [Purdy et al, 1989], have been used. All these direct (with the use of the special stimuli) or indirect techniques (with the use of special signal processing approaches) have been applied to assess the auditory threshold at the speech frequency range and the detection of possible cochlear lesions [Ponton et al, 1992, Philips, 1988, Munnerley et al, 1991].

The purpose of the present work is to extent the use of narrow band tests in the diagnosis of retrocochlear lesions. The analysis of narrow band stimuli ABR in conjunction with wide band stimuli ABR is attempted, taking into account the fact that different frequencies stimulate different regions of the cochlear [Nieuwenhuys et al, 1978, Selters & Brackmann, 1977, Starr & Squires, 1982, Stapells et al 1997, Xu et al 1997, Bunke et al 1998] and are propagated with different groups of nerve fibres along neural pathways which include different intervening nuclei (Fig. 1). The main aim is to combine the already known parameters obtained using clicks with those of the narrow band stimuli ABRs which characterize «frequency specific neural pathways» [Ponton et al, 1992]. It is hoped that these multiparametric comparisons may improve the diagnostic efficacy of ABR giving additional data related to the localization and the extension of the lesion [Kramer & Teas, 1979]. Furthermore, ABR data of clicks and pips from normal subjects as well as from patients with diseases clinically and radiologically are compared.

[pic]

Fig 1: Representation of the tonotopic organization of the auditory nerve (low, medium and high frequencies), along the cochlea (left) the intervening nuclei (center), of the different frequencies (right diagram).

II. MATERIAL AND METHODS

Eighty human subjects participated in the present study; Fifty of them (30 men and 20 women) were adults aged 20-40, with normal hearing and were used as control group; Twenty were patients, aged 25-53 with impaired hearing. This group was divided into two subgroups on the basis of their different response between 250 and 8000 Hz pure-tone thresholds. One subgroup consisted of 15 patients with about 40 dB high-frequency hearing loss (in the frequency band 4000-8000 Hz) and the other included five patients with low-frequency hearing loss (in the frequency band 500-2000 Hz). Ten patients (3 men and 2 women), aged 23-50, selected form hospital population had acoustic neuromas, proven radiologically with CT scans, presenting less or about 40 dB hearing loss (in the frequency band 4000-8000 Hz).

[pic]

Fig. 2: a) The stimulus voltage of a 4000 Hs tone-pip having rise and fall time of 5 ms and 1 ms plateau. b) The amplitude spectrum of the tone-pip. The peak is at 4000 Hz, and the half-power bandwidth is less than 200 Hz.

Following conventional behavioral audiometry, ABR tests were performed in all subjects using auditory stimuli of broad spectrum with an intensity of 80 dB nHL, repetition rate 11.29/sec and frequency filters 100-3000 Hz. Silver disk electrodes were placed as follows: the active on the vertex (Cz), the reference on mastoids (MI) and the “earth” electrode on forehead. Amplified responses were averaged in 2000 trials and analyzed through an on-line computer. Since broad spectrum stimuli evoke brainstem potentials by the synchronous firing of more individual neural units, these potentials are very clear; however a fast-onset stimulus contains transient acoustic energy at various frequencies, which diminishes its frequency specificity. In order to eliminate such transients and obtain a good stimuli frequency specificity, the ABR of the subjects was also tested, using acoustic stimuli with sufficiently long rise time [Starr & Don, 1988, Fausti et al, 1992, Buckard, 1991, Beattie et al 1996, Xu et al, 1997]. We used acoustic stimuli consisting of “tone-pips” with rise time of 5 sinusoidal cycles, plateau 1 and fall 5 cycles. The central frequencies were at 4000 and 8000 Hz and the half-power bandwidth at 200 Hz [Conijn et al, 1992]. The envelopes of their acoustic spectra, plotted on a logarithmic time-base are shown in Fig. 2. The final recordings were analyzed and named in the same way as those of ABR of broad band stimulus [Davis & Hirsh, 1979, Bunke et al 1998].

Using tone pips at 4000 Hz and 8000 Hz the latencies of waves III and V as well as, the interpeak latencies III-V were determined for normals and patients. The latency of wave I as well as the interpeak latencies I-III and I-V were determined only for tone-pips at 8000 Hz. On this basis a multiparametric analysis of the examined pathological cases was performed (Fig. 3).

III. RESULTS

Our results (Table I) show that in normal subjects there is a statistically significant (p ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download