Peter K. Plinkert

Fast detection of wave V in ABRs using a smart single sweep analysis system

The analysis of auditory brainstem responses (ABRs) is accepted to be the most reliable method for the objective diagnosis and quantification of hearing loss in newborns. However, in currently available setups, a large number of sweeps has to be averaged to obtain a meaningful signal at low stimulation levels due to a poor signal-to-noise ratio. In this study, we present a new approach to the detection of wave V in ABRs using a smart single sweep analysis system. A small number of sweeps is decomposed by optimized tight frames and evaluated by a kernel based novelty detection machine. This hybrid supervised learning scheme is combined with an intersweep dissimilarity tracing for the final decision making. At the challenging stimulation level of 30 dB, our system reached a reasonable specificity and sensitivity for the detection of wave V in a fraction of the measurement time of conventional schemes.

Analysis and detection of binaural interaction in auditory evoked brainstem responses by time-scale representations

The beta-wave of the binaural interaction component (BIC) in auditory evoked brainstem responses has been shown to be an objective measure of binaural interaction. However, a reliable and automated detection of this component capable of clinical use still remains a challenge. In this study, wavelet based time-scale representations of auditory evoked brainstem responses were investigated for the analysis of binaural interaction and for an automated detection of the beta-wave. Twenty normal hearing subjects with verified normal directional hearing and speech intelligibility in noise were included in our study. In all of these subjects, the BICs exhibited a characteristic concentration of energy in the time-scale domain which allowed for an automated detection of the beta-wave. Moreover, our study provides an explanation why the beta-wave is hard to detect for larger interaural time delays using time-scale entropy based arguments. It is concluded that time-scale representations of auditory brainstem responses are well suited for the analysis of binaural interaction and allow for an automated detection of the beta-wave.

Phase III results with a totally implantable piezoelectric middle ear implant: Speech audiometry, spatial hearing and psychosocial adjustment

Objective To evaluate the treatment efficacy of an electromechanical middle ear amplifier implant (AI) in patients with chronic moderate-to-severe sensorineural hearing loss (SNHL). The AI is a piezoelectric system with a sound processor and a rechargeable battery within a hermetically sealed titanium canister. Its titanium-sealed microphone is placed in the bony region of the ear canal. The incus-coupled transducer (actuator), which is also inside a titanium casing, is fastened to the adjacent bone. Material and Methods This was a phase III study comprising 20 intention-to-treat patients. Telemetrical adjustments followed electromechanical amplifier implantations. We used a word recognition test as our primary efficacy measure (Freiburg Speech Recognition Test; DIN 45621). Secondary efficacy measures were the sentence comprehension test (Goettinger Satztest, 1996) for auditory orientation within noisy and quiet environments and a psychosocial adjustment test (Gothenburg Profile Test, 1998). The 6-month follow-up comprised a complete medical examination. Nineteen patients completed the study (per-protocol patients; 100% reference). Results Seventeen patients (89%) demonstrated improved binaural recognition of phonetically balanced monosyllables. Fourteen postoperative patients (74%) attained a perfect score (100%) on this test, compared to only 3 preoperative patients (16%). Thirteen patients (68%) reached the sentence recognition threshold at a 2:1 dB signal-to-noise ratio during noisy trials. Correct identification of the noise source direction in the horizontal plane occurred in 89% of the trials. The Gothenburg Profile Test scores showed that the subjective evaluation of hearing, orientation, social behavior and self-confidence increased from 48% to 88%. Three patients did not benefit from the implant. Conclusion Treatment of SNHL with a totally implantable hearing system can be an efficient method for those patients unable to wear hearing aids. However, in order to avoid implantation in non-responders, there is a need for more specific audiological indication criteria.

A time-frequency feature extraction scheme for the automated detection of binaural interaction in auditory brainstem responses

The binaural interaction component (BIC), the difference between the summed monaurally evoked potentials of each ear and the binaurally evoked brainstem potentials, has been shown to be related to directional hearing. However, the detection of the ß-peak as the most consistent part of the BIC is often difficult. Furthermore, there is no clearly defined signal feature characterizing the difference between the monaurally and the binaurally evoked brainstem responses. A closer look at the signals shows that amplitude differences as well as latency differences and variations in wave V slopes could be the reason for the formation of a ß-peak. Using a time-scale feature extraction scheme, we were able to define a signal feature (morphological local discriminant bases ( MLDB) coefficient 1) that accounts for the difference between the sum of the monaurally and binaurally evoked brainstem potentials. With use of this signal feature, reliable automated detection of differences between monaurally and binaurally evoked potentials is possible. As coefficient 1 replicates the behaviour of subjective measurements as well as of the BIC measurements, it can also be seen as a correlate of binaural interaction. With use of this signal feature, it is possible to judge from a given binaurally evoked potential whether it contains information on binaural interaction or not, without comparing it to the sum of the monaurally evoked brainstem responses. Consequently, binaural interaction can be assessed by one, instead of three, measurements by using the method described in this paper. Sumario El componente de interacción binaural (BIC), la diferencia entre la suma de los potenciales evocados monoaurales de cada oído y los potenciales evocados binaurales del tallo cerebral, se ha relacionado con la audición direccional. Sin embargo, la detección del pico ß como la parte más consistente del BIC, es a menudo difícil. Más aún, no existe un rasgo de sen˜al claramente establecido que caracterice la diferencia entre las respuestas evocadas monoaurales y binaurales del tallo cerebral. Una revisión más cercana de las sen˜ales demuestra que las diferencias en la amplitud, tanto como las diferencias en latencia y las variaciones en la pendiente de la onda V, podrían ser la razón para la formación del pico ß. Utilizando un esquema de extracción de rasgos a escala temporal, pudimos definir una rasgo de sen˜al (el coeficiente 1 de bases discriminantes morfológicas locales - MLDB), que representa la diferencia entre la suma de los potenciales evocados monoaurales y binaurales del tallo cerebral. Con el uso de este rasgo de sen˜al, es posible realizar confiablemente la detección automatizada de las diferencias entre los potenciales evocados monoaurales y binaurales. Dado que el coeficiente 1 replica el comportamiento de las medidas subjetivas tanto como las medidas del BIC, también puede ser visto como un elemento de correlación con la interacción binaural. Con el uso de este rasgo de sen˜al, es posible juzgar de un potencial evocados binaural dado, si contiene o no información sobre interacción binaural, sin necesidad de compararlo con la suma de las respuestas evocadas monoaurales del tallo cerebral. Consecuentemente, la interacción binaural puede ser evaluada por medio de una, en lugar de tres mediciones, utilizando el método descrito en este trabajo.

Objective detection of the central auditory processing disorder:A new machine learning approach

The objective detection of binaural interaction is of diagnostic interest for the evaluation of the central auditory processing disorder (CAPD). The /spl beta/-wave of the binaural interaction component in auditory brainstem responses has been suggested as an objective measure of binaural interaction and has been shown to be of diagnostic value in the CAPD diagnosis. However, a reliable and automated detection of the /spl beta/-wave capable of clinical use still remains a challenge. We propose a new machine learning approach to the detection of the CAPD that is based on adapted tight frame decompositions which are tailored for support vector machines with radial kernels. Using shift-invariant scale and morphological features of the binaurally evoked brainstem potentials, our approach provides at least comparable results to the /spl beta/-wave detection in view of the discrimination of subjects being at risk for CAPD and subjects being not at risk for CAPD. Furthermore, as no information from the monaurally evoked potentials is necessary, the measurement cost is reduced by two-thirds compared to the computation of the binaural interaction component. We conclude that a machine learning approach in the form of a hybrid tight frame-support vector classification is effective in the objective detection of the CAPD.

Adapted filter banks in machine learning: applications in biomedical signal processing

The theory of signal-adapted filter banks has been developed in signal compression in recent years and only rarely be applied to other applications fields such as machine learning. In this paper, we propose lattice structure based signal-adapted filter banks and time-scale atoms, respectively, for the construction of morphological local discriminant bases and hybrid wavelet-support vector classifiers. The first mentioned method is a more powerful construction of the recently introduced local discriminant bases algorithm which employs, in addition to the conventional wavelet-packet tree adjustment, an adaptation of the analyzing time-scale atoms. The latter mentioned method utilizes adapted wavelet decompositions which are tailored for support vector classifiers with radial basis functions as kernels. For both methods, we present applications in biomedical signal processing.

TEOAE amplitude growth, detectability, and response threshold in linear and nonlinear mode and in different time windows.

Transitory evoked otoacoustic emissions (TEOAE) have been recorded in 60 ears of 31 adult volunteers with nearly normal hearing at stimulus levels ranging from 83 dB SPL peak equivalent down to the individual response threshold using linear and nonlinear recording mode. The stimulus level dependence of response incidence and amplitude has been analysed for the integral response and in time windows selecting response components of limited latency ranges. At stimulus levels above 70 dB SPL peak equivalent the TEOAE records received in linear mode are contaminated with stimulus artifacts. At moderate stimulus levels the TEOAE amplitude differs only to a small extent between the two recording modes. At low levels the linear mode turns out to be better suited for signal detection due to its inherent lower noise level. The response threshold, defined as the highest stimulus level yielding a reproducibility of at least 60%, is significantly correlated to hearing threshold. The consideration of time windowed responses yields best results with respect to incidence and threshold of responses in the latency range between 5 and 10 ms, but it does not enhance frequency specificity.

Hybrid wavelet-kernel based classifiers and novelty detectors in biosignal processing

The recognition of waveforms represents a major challenge in biosignal processing. In this area, the recognition scheme has to offer a particular high generalization performance and should often allow for the inclusion prior knowledge about the waveforms. In this paper, we propose a hybrid machine learning scheme as general approach to waveform recognition in the biomedical area. Our hybrid scheme is based on feature extractions by adapted filter banks and support vector machines or kernel based novelty detectors. It allows for the inclusion of a priori knowledge such as local instabilities in time and shift-variance of bioelectric waveforms. We apply our scheme for the classification of endocardial waveforms and the detection of transient evoked otoacoustic emissions. For both applications, we show that our scheme outperforms conventional methods used before.

Contact endoscopic findings in hereditary hemorrhagic telangiectasia

Hereditary hemorrhagic telangiectasia is characterized by angiodysplastic lesions. So far, knowledge is limited on the vascular architecture and rate of occult manifestation of telangiectases. Contact endoscopy has not been used for this task before.

HNO-Notfälle – Die wichtigsten Erkrankungen im Überblick

Notfallsituationen im Hals-Nasen-Ohren-Bereich sind relativ haufig. Um eine vitale Bedrohung des Patienten bzw. schwere bleibende Schaden abzuwenden, mussen Sie als Arzt schnell handeln – auch wenn Sie kein HNO-Experte sind. Orientierung geben die auffalligen Leitsymptome, wie z. B. Atemnot oder Blutungen. Sie grenzen die Diagnose der zugrundeliegenden Erkrankung stark ein und erleichtern die rasche Auswahl der passenden Soforttherapie. Diese Ubersichtsarbeit stellt die wichtigsten Notfalle an Hals, Nase und Ohren vor und skizziert, wie man Patienten optimal versorgt.