Multiresolution wavelet analysis of evoked potentials

Abstract

Neurological injury, such as from cerebral hypoxia, appears to cause complex changes in the shape of evoked potential (EP) signals. To characterize such changes we analyze EP signals with the aid of scaling functions called wavelets. In particular, we consider multiresolution wavelets that are a family of orthonormal functions. In the time domain, the multiresolution wavelets analyze EP signals at coarse or successively greater levels of temporal detail. In the frequency domain, the multiresolution wavelets resolve the EP signal into independent spectral bands. In an experimental demonstration of the method, somatosensory EP signals recorded during cerebral hypoxia in anesthetized cats are analyzed. Results obtained by multiresolution wavelet analysis are compared with conventional time-domain analysis and Fourier series expansions of the same signals. Multiresolution wavelet analysis appears to be a different, sensitive way to analyze EP signal features and to follow the EP signal trends in neurologic injury. Two characteristics appear to be of diagnostic value: the detail component of the MRW displays an early and a more rapid decline in response to hypoxic injury while the coarse component displays an earlier recovery upon reoxygenation.

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

“…Among these techniques, the wavelet transform provides a good time resolution at high frequencies and a good frequency resolution at low frequencies [Rioul and Vetterli, 1991], and it finds a rapidly growing number of applications in the biomedical signal processing field and, particularly, in evoked potential analysis [Thakor et al, 1993;Bertrand et al, 1994;Raz and Turetsky, 1995;Samar et al, 1995].…”

Over-Complete Discrete Wavelet Transformation of the Normal Auditory Brainstem Response Improves Prediction of Outcome following Severe Acute Closed Head Injury

“…Such a possibility is consistent with similar reports that: mid-frequency ABR FFT component magnitudes (at 500 Hz) were smaller in severe ACHI subjects who lived compared to those who died [Hall, 1986]; high-frequency C3 wave magnitudes (reconstructed from the 700-to 1200-Hz content of their ABR FFT) were smaller in spinocerebellar degeneration subjects compared to normal subjects [Yokoyama et al, 1994]; high-frequency ABR waveforms (reconstructed from the 700-to 1200-Hz content of their ABR FFT) were the most sensitive to lesions of the auditory brainstem in cats ; high-frequency somatosensory evoked potentials (reconstructed from a high-frequency DWT D4 scale; frequency range not stated) were the most sensitive to the onset of hypoxia in cats [Thakor et al, 1993]; and high-frequency somatosensory evoked potential energy (reconstructed in the 50-to 275-Hz range using timefrequency windowing) deteriorated an average of 2.75 min earlier than the normal somatosensory evoked potential signal in anaesthetized cats with induced hypoxia [Braun et al, 1996]. As some authors have postulated that the high-frequency components of the evoked potentials (including the ABR) represent a separate 'faster fi ring' population of nerve fi bres [Hall, 1992], it would appear that these 'fast fi ring' nerve fi bres could be more susceptible to damage from severe ACHI and other neurological insults.…”

Estimation of Latency Changes and Relative Amplitudes in Somatosensory Evoked Potentials Using Wavelets and Regression

“…The application of wavelet transforms in biomedical systems (20) and in evoked potentials (21) has been reviewed. It has been shown that the wavelet transform, giving information in the time/frequency domains, is superior to conventional time series analysis and Fourier transforms (22,23). Some applications of wavelet transforms in biomedical systems were reported, such as: uterine EHG signal denoising (24), evoked potential reconstruction (25,26), EP amplitude estimation (23), EP latency tracking (10), EP source localization (27,28), EEG seizure detection (29), and feature extraction for medical diagnosis (30,31).…”