Purchase this article with an account.
Thomas Meigen; Trend Artifacts Can Be Removed From Electrophysiological Recordings Within The Fourier Domain. Invest. Ophthalmol. Vis. Sci. 2011;52(14):6092.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Electrophysiological ERG and VEP recordings may contain residual trend artifacts induced by eye movements or blinks. These trend artifacts impair the resulting amplitude values. The traditional solution is (1) to calculate the difference between the first and last data point, and (2) to subtract a linear function with the same voltage difference. This solution, however, fails to predict correct steady-state signals when noise intrusions with a non-integer number of cycles from mains interference are present, which may also add to the difference between the first and last data point. The aim of the study was to present a method to remove both artifacts in one step by an appropriate matrix multiplication in the Fourier domain and to validate this method with Monte Carlo simulations. By removing both noise components in one step no detrimental side-effects are introduced compared to the traditional strategy to remove both artifacts one-by-one.
A sinusoidal waveform with an amplitude of 4 µV and a frequency of 6.7 Hz defined the steady-state test signal within a time interval of 1,05 s. A linear trend, a gaussian background noise component, and intrusions from mains interference were added with random amplitude, while the mean values of these noise components were modified parametrically. For each parameter set, 50 averages of 50 trials were calculated. The distance between the averaged waveform and the test signal was quantified in the complex Fourier domain for 3 signals. These signals were (a) the average of the trials without trend removal, (b) a waveform after subtraction of an appropriate linear ramp from (a), and (c) a waveform after back transformation of the Fourier spectrum from (a) where all components from linear trend and mains interference were removed after considering their respective overlap in the Fourier domain.
(1) Both methods (b) and (c) showed significantly smaller deviations from the test signal than the original waveform (a). (2) The temporal gain for (b) and (c) to reach the same distance from the test signal as (a) was proportional to the amplitude of the linear trend artifact and inversely proportional to the magnitude of the background noise. (3) While the distances of (a) and (b) from the test signal increased with increasing amount of mains interference, the quality of the waveform (c) was independent of this noise intrusion.
The new methods allows to reliably remove trend artifacts from electrophysiological recordings irrespective of additional intrusions from mains interference and may help to improve the post-processing of electrophysiological recordings.
This PDF is available to Subscribers Only