Instantaneous Energy Spectrum Analysis For Frequency Following Response Of Speech Evoked Brainstem Response

An auditory evoked potential (or response)(AEP) is an electrical potential recorded from the auditory nervous system of a human of other animal following presentation of an auditory stimulus. AEP is can be used to trace the signal generated by a sound through the ascending auditory pathway. Discovered nearly40years ago, the auditory brainstem response (ABR) is an AEP generated by the synchronous activity of populations of neurons in the brainstem and recorded via electrodes placed on the scalp. The resulting recording is series vertex positive waves labeled with Roman numerals in Jewett and Williston convention, Ⅰ～Ⅶ. and occurring in the first10ms after onset of an auditory stimulus. The ABR to simple stimuli such as clicks and sinusoidal tones are widely used in clinical practice as a metric for determining auditory thresholds, detecting neuropathologies and the evaluation of auditory pathway integrity. However, these simple stimuli are poor approximations of the behaviorally relevant complex sounds that we encounter outside the laboratory (e.g., speech and music, non-speech vocal sounds, and environmental sounds).The human soundscape is characterized by complex sounds with rich harmonic structures, dynamic amplitude modulations, and rapid spectrotemporal fluctuations. This complexity is represented by an exceptionally precise temporal and spectral neural code within the auditory brainstem, two broad classes of time-locked responses can be defined, namely, transient and sustained. As the names suggest, brief, nonperiodic stimulus features evoke transient responses, whereas periodic features elicit sustained phase-locked responses. Although clicks and tones have been instrumental in defining these basic these transient and sustained responses patterns and complex sounds include both transient and sustained features, the response to complex sound is not necessarily predictable from the response to click and tones. For these reasons, auditory neuroscience has gradually transitioned to using sounds that are more complex. One of the most extensively studied is ABR to speech syllable (speech-ABR) such as/da/. It should be pointed out that we use the term "ABR" to describe both transient and sustained responses originating in the auditory brainstem.40ms speech syllable/da/contains transient and sustained portions. The former is stop consonant/d/which lasts10ms. The latter is vowel/a/with periodic patterns lasting30ms, in which there are three quasi-periodic waves denoted by wave d, e and f, respectively. The period of these three quasi-periodic waves is about10ms. And the vowel contains information at fundamental frequency (F0) and five formant frequencies (F1-F5). Fo is103-121Hz, F, is220-720Hz, F2is1700-1240Hz, F3is2580-2500Hz, and F4,5is3600-4500Hz.Because speech-ABRs provide direct information about how the sound structure of a speech syllable is encoded by auditory system and it is particularly compelling to consider that specific aspects of the sound structure of the acoustic signal are maintained and reflected in the neural code, similar to speech syllable itself, speech-ABR can be divided into transient and sustained portions, namely the onset response (OR) and the frequency following response (FFR). OR is induced by stop consonant/d/and contains wave V and wave A, with peaks durations lasting tenths of milliseconds, thus we will refer to these rapid deflections as transient responses. We use the term "FFR" to refer the later portion of the response evoked by the harmonic vowel structure of the speech syllable/da/. Within the FFR are discrete peaks corresponding to the periodic peaks in the stimulus waveform. Three periodic peaks D, E and F in FFR represent how the brain stem encodes the fundamental frequency of vowel, corresponding to the periodic peaks in vowel/a/waveform. Some researches revealed that waves D, E and F are significantly correlated with each other, which FFR may be considered as a general physiological index and originates the same neural processing mechanism. And the temporal periodicity of FFR contains important information needed for speech perception. Therefore, the integrity of periodic peaks D, E and F could be applied to evaluate the quality of encoding of vowel in brainstem. However, in practical application, FFR in individual speech-ABR to syllable/da/is often difficult to identify by visual inspection, even when FFR in grand averaged waveform of speech-ABR to syllable/da/is obvious, which makes difficult to decide whether the vowel is encoded in brainstem in clinical practice.Speech-ABR is nonlinear and non-stationary because of nonlinear nature of the auditory system. A recently developed method for analyzing nonlinear and non-stationary data is Hilbert-Huang transform (HHT). The key part of HHT is called empirical mode decomposition (EMD), with which any complicated data can be decomposed into a finite and often small number of intrinsic mode function (IMF). The IMFs can be heuristically interpreted by spectral arguments, but unlike other time-frequency analysis method, such as wavelet transform, the IMFs are not separated totally based on the frequency elements it contains. Any peak in the speech-ABR waveform may contain the first harmonic and several another harmonics in time-domain. After performing EMD on the individual speech-ABR, each IMF obtained contains only one frequency at any time point. FFR reflects phase-locked activity elicited by periodic acoustic features. If FFR exists, waves D, E and F might mainly centralized on the same layer of IMF.In this study, the idea of the defined instantaneous energy (IE) is based on HHT, in which the key decomposing criterion is, we believe, analogous to the organizing way of the responses.The rationale of IE description for FFRs is that according to the fidelity attribute of speech-ABR to the acoustic structure of the stimulus. A properly defined IE description might be better to comprehend the FFR properties than the conventional way that is amplitude-oriented.The specific steps of the method are as follows: (1) Individual speech-ABR is decomposed into several IMFs using EMD;(2) Computethe instantaneous energy of each IMF based on Hilbert transform;(3) Observe the instantaneous energy spectrum of IMFs and record the layersindexes where the waves D, E and F appear;(4) Detect waves D, E and F in three separate searching windows of10ms centered at{17-27,27-37,37-47}ms, and write down their latencies and polarities.There are29adults passing a simple hearing test participated in speech-ABR experiment.The analysis procedures are:I)Candidate selection. Experts from the relevant expertise are invited to visually screen the data in order to select speech-ABRs with distinct wave V (indicating acceptable recordings) as the candidates for further analysis; andfour subjects were rejected. II)FFR screening on raw speech-ABRs. Accordingly,28%(7cases out of25) were classified as distinct in the appearance of FFRs. Ⅲ) FFR screening based on IE spectra. IE spectra of each IMF by EMD were calculated;23(92%) speech-ABRs were found to be distinct in IE spectra.The results show that this method can represent the major properties of FFR in the speech-ABR efficiently. The proposed method has two significant advantages compared with the conventional one:I) FFR in instantaneous spectrum is significantly more obvious than in individual speech-ABR temporal waveform; II) not all of the polarities of waves D, E and F in individual speech-ABR temporal waveform are same, but they are all positivity waves in instantaneous spectrum. It thus can conclude that the IE method can be promising in clinic application for providing a new way of representing FFR in individual speech-ABR.