Adaptive Denoising and Multiscale Detection of the V Wave in Brainstem Auditory Evoked Potentials

Abstract

This paper describes a wavelet-transform-based system for the V wave identification in brainstem auditory evoked potentials (BAEP). The system combines signal denoising and rule-based localization modules. The signal denoising module has the potential of effective noise reduction after signal averaging. It analyses adaptively the evolution of the wavelet transform maxima across scales. The singularities of the signal create wavelet maxima with different properties from those of the induced noise. A non-linear filtering process implemented with a neural network extracts out the noise-induced maxima. The filtered wavelet details are subsequently analysed by the rule-based localization module for the automatic identification of the V wave. In the first phase, it implements a set of statistical observations as well as heuristic criteria used by human experts in order to classify the IV–V complex. At the second phase, using a multiscale focusing algorithm, the IV and V waves are positioned on the BAEP signal. Our experiments revealed that the system provides accurate results even for signals exhibiting unclear IV–V complexes.

Automated Analysis of the Auditory Brainstem Response Using Derivative Estimation Wavelets

“…Methods previously used for analysis of ABR peaks in suprathreshold waveforms have included: detecting the local maxima in low-pass fi ltered waveforms [Pratt et al, 1989], designing fi lters matched to the peaks of interest [Vannier et al, 2001;Vannier et al, 2002;Woodworth et al, 1983] and applying artifi cial neural networks (ANNs) [Alpsan et al, 1994;Popescu et al, 1999;Tian et al, 1997]. However, the most commonly used method has been to band-limit the ABR waveform and then to estimate its gradient (fi rst derivative).…”

“…However, the most commonly used method has been to band-limit the ABR waveform and then to estimate its gradient (fi rst derivative). The peaks and troughs of interest are then found via the zero crossings in the gradient waveform (with the peaks being distinguished from the troughs by the direction of the zero crossing) [Boston, 1989;Delgado and Ozdamar, 1994;Fridman et al, 1982;Gabriel et al, 1980;Pool and Finitzo, 1989;Popescu et al, 1999]. Once candidate zero crossings are found, individual peaks can be labelled according to expected latency and amplitude criteria [Gabriel et al, 1980;Pool and Finitzo, 1989], estimates of the amount of noise present [Boston, 1989] or previous knowledge of peak latencies from higher stimulus intensity waveforms [Vannier et al, 2002].…”

“…More accurate estimates of the peak latency can then be made using a slightly broader band-pass fi lter centred on 500 Hz [Pratt et al, 1989;Delgado and Ozdamar, 1994;Popescu et al, 1999]. This coarse-to-fi ne [Mallat, 1999] strategy improves detection at low stimulus intensities as the narrowband fi lter has an improved SNR, whilst the broader band fi lter has improved temporal resolution [Marr and Hildreth, 1980].…”

“…Whilst many previous systems have used various multiresolution detection methods, only one [Popescu et al, 1999] has formalised this process using wavelet analysis [Daubechies, 1992;Mallat, 1999]. Because of this, the majority of the previous systems can be criticised for being potentially ad hoc and therefore suboptimal.…”

“…Another limitation of many of the previous automated wave labelling systems has been that they have only attempted to identify the primary peaks of the ABR (I, III and V) [Pool and Finitzo, 1989;Pratt et al, 1989] or alternatively the primary IV-V complex [Popescu et al, 1999]. None of them attempts to identify waves II, IV, VI or VII.…”