Preliminary Investigation Of The Linear/Nonlinear Components In Maximum Length Sequence Evoked Auditory Brainstem Response

The auditory brainstem response (ABR) reflects bioelectrical potential arising from the auditory nerve and brainstem in response to acoustic short-duration sounds such as a click. It typically consists of a series of peaks appearing within the first 10 ms after the onset of a stimulus. These peaks are labeled with Roman numerals Ⅰ-Ⅶ, in which the wave Ⅰ,Ⅲ,Ⅴ are more stable and distinct than others, and wave V is the most stable component. Therefore, the latency and amplitude of wave V, the interpeak latencies among wave Ⅰ,Ⅲ,Ⅴ, the ratio of wave I and wave V amplitude et al are analysized to make localization diagnosis for the lesions of auditory nerve pathways.The conventional approach for acquiring ABR (cABR) requires linear averaging of the response to every stimulus to attuenatte the random noises and artifacts. The stimulus rate for cABR is limited due to averaging. If the stimulus rate is any higher, distorted response will occur due to overlapping of the transient ABRs. Eysholdt and Schreiner reported a deconvolution method using maximum length sequence (MLS) technique to restore the traisent ABR at high stimulus rates. This method assumes a linear convolution between the ABR and MLS. This meothd can be extended to a more general form which is able to deal with both linear and nonlinear cases as proposed by Shi and Hecox, using cross-correlation technique, under this way,the ABR is called MLS-ABR which featured waves were labled by Arab numbers. A nonlinear system could be described and analyzed with Volterra series presentation, in which the Volterra kernel’slices describe the effect of interaction among the input pulses on the observed responses. When m-pulse sequence, a variant of MLS, is the input of this system, the Volterra kernel slices (VS) could be obtained by the cross-correlation function between the output and recovery sequence. The linear component of MLS-ABR is similar to the cABR, which is an adapted responses to a isolated stimulus. The nonlinear components of MLS-ABRs are the activiety in response to the temporal interaction among multiple stimuli.To date, the knowledge of MLS-ABR is far from enough, this study aims at a prilimiary investigation on the fundamental characteristics of MLS-ABR. A precausion needs to be made in the selection of MLS, kernel slices distortion due to overlap may occur for inappropriate MLS. In this paper, we presented a selection criteria of MLS, and determined a 9th order MLS accordingly.In the ABR experiments, we recorded both MLS-ABR and cABR from 16 healthy adults, and carried out comparison study on the morphology and time-domain parameters of the ABRs. The results demonstrated a salient first-order and second-order kernel slices (VS1 and VS21) with high elicitations (>80%) of the selected featured waves. As expected, the wave 5 exhibited strong nonlinear effects. The variable coefficients were used to quantify the latency stability of the featured waves in comparison with that of cABRs. The results presented that the nonlinear ABR is more stable than linear ABR. For the linear ABRs, the stabilities of wave 1/Ⅰ, 3/Ⅲ are similarity (the variable coefficients of wave 1,3 are 6.55,4.82, respectively; and the variable coefficients of wave Ⅰ,Ⅲ are 6.56,4.76, respectively), but the wave 5 of the linear component in MLS-ABR (the variable coefficient of linear component in MLS-ABR is 2.37) is more stable than cABR (the variable coefficient of cABR is 3.64). This study justifies the presence of nonlinearity of ABR, and demonstrates the fundamental characteristics of MLS-ABRs.The cABR has been wildly used in the clinical audiometry. Because the different mechanisms of the neuronal processing at different sites, different synchronization timing of the stimulus information, which directly affects the amplitude and latency of ABRs, it is thus high desired to characterize the featured components by peak latencies and interpeak latencies. In this study, we also attempted to explore the feasibility of MLS-ABRs in audiometry so that correlation analysis was conducted between featured components of cABRs and MLS-ABRs. Ten subjects with elicited waves of Ⅰ/1,Ⅲ/3 and V/5 was selected for this comparison. The correlation coefficients of peak latency between the linear component of Ⅰ/1, Ⅲ/3, and Ⅴ/5 are 0.81,0.73 and 0.64, respectively; likewise, the correlation coefficients of peak latencies between the linear/nonlinear components of waves Ⅰ/1 and Ⅴ/5 are 0.76 and 0.67, respectively; the correlation coefficients of peak latency between the linear and nonlinear component of waves 1 and 5 are 0.87 and 0.82, respectively. The correlation coefficients of interpeak latencies between the waves Ⅰ/1～Ⅲ/3, Ⅲ/3～Ⅴ/5, Ⅰ/1～Ⅴ/5 of cABR and MLS nonlinear componet are 0.68,0.75 and 0.84, respectively; the correlation coefficients of interpeak latency between waves Ⅰ/1～Ⅴ/5 of cABR and the nonlinear componet is 0.73; and the correlation coefficients of interpeak latencies between the waves 1-5 is 0.78. No significant correlation was found between ABR waves Ⅲ/3. All these results indicate the homogenesis nature of linear and nonlinear components.In summary, this paper introduces the theory and technology of extracting nonlinear response, describes the fundamental characteristics of MLS-ABR, and compares it with the characteristics of the well-established cABR. This preliminary study provides further insight into the brainstem activity, suggesting valuable clinic implication of the devoleoping MLS-ABR.