An automatic algorithm for blink-artifact suppression based on iterative template matching: Application to single channel recording of cortical auditory evoked potentials
Date
2018Author
Valderrama, Joaquin T
de la Torre, Angel
Van Dun, Bram
Metadata
Show full item recordAbstract
Artifact reduction in electroencephalogram (EEG) signals is usually
necessary to carry out data analysis appropriately. Despite the large amount of
denoising techniques available with a multichannel setup, there is a lack of e cient
algorithms that remove (not only detect) blink-artifacts from a single channel EEG,
which is of interest in many clinical and research applications. This paper describes
and evaluates the Iterative Template Matching and Suppression (ITMS), a new method
proposed for detecting and suppressing the artifact associated with the blink activity
from a single channel EEG.
Approach: The approach of ITMS consists of (a) an iterative process in which
blink-events are detected and the blink-artifact waveform of the analyzed subject is
estimated, (b) generation of a signal modeling the blink-artifact, and (c) suppression
of this signal from the raw EEG. The performance of ITMS is compared with the
Multi-window Summation of Derivatives within a Window (MSDW) technique using
both synthesized and real EEG data.
Main results: Results suggest that ITMS presents an adequate performance in
detecting and suppressing blink-artifacts from a single channel EEG. When applied
to the analysis of cortical auditory evoked potentials (CAEPs), ITMS provides a
signi cant quality improvement in the resulting responses, i.e. in a cohort of 30 adults,
the mean correlation coe cient improved from 0.37 to 0.65 when the blink-artifacts
were detected and suppressed by ITMS.
Signi cance: ITMS is an e cient solution to the problem of denoising blinkartifacts
in single-channel EEG applications, both in clinical and research elds.
The proposed ITMS algorithm is stable; automatic, since it does not require human
intervention; low-invasive, because the EEG segments not contaminated by blinkartifacts
remain unaltered; and easy to implement, as can be observed in the Matlab
script implemeting the algorithm provided as supporting material.