Does filtering and smoothing of average evoked potentials really pay? A statistical comparison.

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Averaging sweeps to obtain evoked potentials (EPs) inadequately reduces background activity with few stimuli.
• Existing methods like a posteriori Wiener filtering, time-varying filtering, and smoothing have been proposed to address this limitation.
• A controversy exists regarding the efficacy of a posteriori Wiener filtering.

Purpose of the Study:

• To statistically compare a posteriori Wiener filtering, time-varying filtering, and smoothing for evoked potential analysis.
• To evaluate the effectiveness of these methods in reducing background activity and preserving signal integrity.
• To assess potential artifacts introduced by filtering and smoothing techniques.

Main Methods:

• Statistical comparison of filtering and smoothing methods using real data from two subject groups.
• Analysis of flash evoked potentials (41 subjects) and pattern reversal evoked potentials (9 subjects).
• Evaluation of signal attenuation, smoothness improvement, interindividual variability, and topographical distribution bias.

Main Results:

• Filtering approaches primarily improved signal by attenuation, not smoothness, with effects dependent on signal-to-noise ratio.
• Filtering introduced artificial increases in interindividual variability and biased topographical distributions.
• Smoothing did not exhibit these biases but caused significant distortions with strong application.

Conclusions:

• Standard filtering methods for evoked potentials can introduce significant artifacts, including signal attenuation and biased topographical data.
• Smoothing offers an alternative but risks substantial signal distortion if applied too strongly.
• Careful consideration of signal-to-noise ratio and potential artifacts is crucial when selecting and applying signal processing techniques for evoked potentials.