Hellmut von Specht

Sorted averaging—application to auditory event-related responses

To isolate event-related potentials (ERPs) from the biological background noise, averaging of stimulus-locked electroencephalogram (EEG) epochs is needed. Recordings with patients often reveal a noisy background EEG, i.e., a high amplitude and multiple artifacts. Research studies have to deal with a limited number of available epochs. Therefore, averaging and efficient artifact rejection algorithms are needed. This paper focuses on the sorted averaging algorithm which was developed for the recording of auditory brainstem responses. We demonstrate the applicability of sorted averaging on ERPs by comparing three different averaging algorithms on a classical mismatch negativity (MMN) paradigm, recorded at 10 normal hearing volunteers. The resulting estimated signal-to-noise ratio (SNR) of the ERP waveforms was significantly increased compared to established averaging algorithms. Thus, the sorted averaging algorithm provides an improvement of the SNR in recordings with poor SNR (e.g., the MMN) or noisy background EEG (e.g., at Cochlear Implant users).

Comparison of Absolute Thresholds Derived from an Adaptive Forced-Choice Procedure and from Reaction Probabilities and Reaction Times in a Simple Reaction Time Paradigm

An understanding of the auditory systems operation requires knowledge of the mechanisms underlying thresholds. In this work we compare detection thresholds obtained with a three-interval-three-alternative forced-choice paradigm with reaction thresholds extracted from both reaction probabilities (RP) and reaction times (RT) in a simple RT paradigm from the same listeners under otherwise nearly identical experimental conditions. Detection thresholds, RP, and RT to auditory stimuli exhibited substantial variation from session to session. Most of the intersession variation in RP and RT could be accounted for by intersession variation in a listeners absolute sensitivity. The reaction thresholds extracted from RP were very similar, if not identical, to those extracted from RT. On the other hand, reaction thresholds were always higher than detection thresholds. The difference between the two thresholds can be considered as the additional amount of evidence required by each listener to react to a stimulus in an unforced design on top of that necessary for detection in the forced-choice design. This difference is inversely related to the listeners probability of producing false alarms. We found that RT, once corrected for some irreducible minimum RT, reflects the time at which a given stimulus reaches the listeners reaction threshold. This suggests that the relationships between simple RT and loudness (reported in the literature) are probably caused by a tight relationship between temporal summation at threshold and temporal summation of loudness.

Amplitude modulation following responses in awake and sleeping humans--a comparison for 40 Hz and 80 Hz modulation frequency.

There have been several studies, which suggest that the amplitude of amplitude modulation following responses (AMFR) is correlated to the state of vigilance, similar to the 40 Hz event-related potentials. The aim of the present study was to compare the dependency of the AMFR-amplitude from the state of sleep for 40 Hz and 80 Hz modulation frequency. Eight normal hearing adults were investigated during natural and drug-induced sleep. The stimuli used were sinusoidally amplitude-modulated tones of 1 kHz carried frequency and 40 or 80 Hz modulation frequency at 60 dB nH stimulation level. For 40 Hz modulation frequency an increase of EEG-activity in the Delta-and Theta-band during periods of sleep correlates significantly with a decreased AMFR-amplitude whereas for 80 Hz no significant relation between stage of sleep and AMFR-amplitude could be found. The results suggest that in audiological use of 40 Hz-AMFR the state of vigilance should be monitored and stabilized at a high level.

Sorted averaging-principle and application to auditory brainstem responses

A new averaging method for evoked potentials, called sorted averaging, is presented. The method requires an ensemble of sweeps stored in memory and is based on the principle of interchangeability of individual sweeps within this ensemble. Sorted averaging is applied by sorting all sweeps according to their estimated background noise and successive averaging of this sorted ensemble, starting with low-noise sweeps. Signal-to-noise power ratio (SNR2) is estimated by calculating the inverse single point variance (Elberling & Don, 1984). This SNR2 estimate increases linearly with the number of sweeps for standard averaging and shows a clear maximum for sorted averaging. Auditory brainstem responses to click stimuli at 70 dB nHL in 20 normal hearing subjects were recorded and 4000 individual sweeps during each run were stored for analysis. In an off-line analysis, SNR2 for standard averaging with 10 microV artefact rejection, for weighted averaging and for sorted averaging were calculated. Sorted averaging was found to yield a significantly higher SNR2.

Near-threshold recordings of amplitude modulation following responses (AMFR) in children of different ages

Amplitude modulation following responses (AMFR) to single near-threshold 40-Hz or 80-Hz amplitudemodulated tones of 1 kHz were recorded in 48 normal children between the ages of 2 months and 14 years. Children under the age of 2 years were tested during natural sleep, and older children were tested awake. The objectives of this study were to find out how the AMFR changes with age and to determine the most effective modulation frequency for objective threshold assessment at different ages. The optimal modulation frequency changed from higher to lower modulation frequencies at about 13 years. In younger children, the 80-Hz AMFR was larger than the 40-Hz AMFR. The 40-Hz response became similar to the adult response by the age of 14 years, while the 80-Hz response had an amplitude greater than half of an adult response by the age of 1 year, changed very little during the investigated period, and reached the amplitude of adult responses at the end of the investigated period. Sumario En 48 nin˜os normales de 2 meses a14 an˜os de edad, se registraron las respuestas consecutivas a modulacio´n de amplitud (AMFR) a tonos u´nicos de 1 kHz casi-umbral de amplitud modulada a 40-80 Hz. Los nin˜os menores a dos an˜os fueron examinados mientras dormi´an naturalmente y los mayores estando despiertos. Los objetivos de este estudio fueron: conocer la forma en que las AMFR cambian con la edad y obtener la frecuencia de modulacio´n ma´s efectiva para la determinacio´n del umbral objetivo a diferentes edades. La frecuencia o´ptima de modulacio´n se modifico´ de alta a baja cerca de los 13 an˜os. En nin˜os menores, la AMFR a 80 Hz fue mayor que la AMFR a 40 Hz. La respuesta a 40 Hz mostro´ similitud a la del adulto a la edad de 14 an˜os, mientras que la respuesta a 80 Hz mostro´ una amplitud mayor que la mitad de la respuesta del adulto al an˜o de edad, cambio´ muy poco durante el tiempo investigado y alcanzo´ la amplitud de la respuesta del adulto al final del periodo de la investigacio´n.

Analysis of ceiling effects occurring with speech recognition tests in adult cochlear-implanted patients.

This article presents a simple method of analysing speech test scores which are biased through ceiling effects. Eighty postlingually deafened adults implanted with a MED-EL COMBI 40/40+ cochlear implant (CI) were administered a numbers test and a sentence test at initial device activation and at 1, 3, 6, 12 and 24 months thereafter. As a measure for speech recognition performance, the number of patients who scored at the ‘ceiling level’ (i.e. at least 95% correct answers) was counted at each test interval. Results showed a quick increase in this number soon after device activation as well as a continuous improvement over time (numbers test: 1 month: 51%; 6 months: 73%; 24 months: 88%; sentence test: 1 month: 33%; 6 months: 49%; 24 months: 64%). The new method allows for the detection of speech recognition progress in CI patient samples even at late test intervals, where improvement curves based on averaged scores are usually assuming a flat shape.

On the frequency spectrum of Amplitude Modulation Following Responses

Objective detection of Amplitude Modulation Following Responses (AMFR) is based on statistics applied after signal transformation from the time to the frequency domain by means of Discrete Fourier Transformation. In theory the frequency resolution of such transformation depends only on the analysed time window. In practise frequency resolution is also limited by the error caused by minimal difference between the clocks used for stimulus generation and Analogue/Digital-conversion. Small differences in clock frequencies may cause a spread of energy to neighbouring bins. In order to avoid this error we derived the sample clock for the A/D-conversion from the stimulator clock. By means of this technique the frequency structure of the AMFR was investigated. It is shown that if technical induced errors are excluded, the energy of the AMFR-response is limited to a very narrow frequency band. No physiologically induced disturbances of the phase locking of the AMFR to the modulation frequency could be observed. Additionally it is demonstrated that an increase of frequency resolution leads to an improved signal to noise ratio similar to the increase of averages in the time domain.

Influence of Electrode Position on Near-Threshold Recording of Auditory Evoked Brainstem Potentials

Use of brainstem potentials in audiology is based on identifying a response in the near-threshold range. Wave V has turned out to be the component of the brainstem potentials that can be detected with a great deal of reliability when assessment of the threshold for the particular stimulus is desirable. As the detection of Wave I can be dispensed within audiologic diagnosis, the question arises as to whether or not a non-cephalic reference electrode compared to the common lateral position of the reference electrode on the ipsilateral mastoid produces a more stable Wave V of increased amplitude. In a group of 20 normal-hearing adults, near-threshold stimulation with clicks was conducted and the brainstem potentials recorded simultaneously with the reference electrode placed in four different locations (ipsilateral mastoid, contralateral mastoid, non-cephalic electrode, and ipsilateral earlobe). At all of the three intensities studied (10 dB nHL, 20 dB nHL and 30 dB nHL), recording with the non-cephalic reference yielded the highest amplitudes, the finding being statistically significant.

Visualization of stimulation patterns in cochlear implants: Application to event-related potentials (P300) in cochlear implant users

Auditory P300 potentials obtained in cochlear implant users were evoked with tone bursts designed to reflect the frequency stimulation patterns of intracochlear electrodes. To visualize these stimulation patterns in MED-EL COMBI 40+ cochlear implants, we calculated color-coded plots of the charge of each stimulus pulse as a function of time and stimulation channel (stimulograms). The aim of this study was to investigate the effect of stimulation patterns on event-related potentials, such as the P300. The influence of electrode separation on the P300 response in postlingually deaf adults using a cochlear implant is demonstrated in two examples.

Reduction of Background Noise in Human Auditory Brainstem Response by Means of Classified Averaging

The most widely used method for the recording of evoked potentials (EP) is the classical technique of ensemble averaging in the time domain. The use of averaging calls for a number of requirements to be met (random noise, independence of noise and signal, deterministic signal in all sweeps). If these requirements are satisfied and the noise is stationary throughout the period of examination, the well-known improvement of the signal-to-noise ratio (SNR) with the square root of the number of sweeps is obtained. The stationarity is frequently violated by either muscular activity (artifacts) or slow changes of the subjects state of relaxation. Two methods are used to minimize the undesirable effects of noise instability on the SNR: The first and most common method used is artifact rejection: Elimination of realizations with signal amplitudes exceeding a certain level from the averaging process. The second method used is weighted averaging: Weighting of realizations or blocks of realizations inversely to the estimated power of background noise (Hoke et al. 1984). Optimal artifact rejection requires a prior knowledge of the noise distribution. Unavoidable underestimation of the signal amplitude is the main disadvantage of weighted averaging (Lutkenhoner et al. 1985). In this paper classified averaging will be presented as an alternative approach to optimal reduction of residual background noise.