Denoising of single-trial matrix representations using 2D nonlinear diffusion filtering

In this paper we present a novel application of denoising by means of nonlinear diffusion filters (NDFs). NDFs have been successfully applied for image processing and computer vision areas, particularly in image denoising, smoothing, segmentation, and restoration. We apply two types of NDFs for the...

Full description

Saved in:
Bibliographic Details
Published in:Journal of neuroscience methods 2010-01, Vol.185 (2), p.284-292
Main Authors: Mustaffa, I., Trenado, C., Schwerdtfeger, K., Strauss, D.J.
Format: Article
Language:English
Subjects:
Algorithms
Brain - physiology
Computer Simulation
Denoising
Diffusion
Electroencephalography
Evoked Potentials - physiology
Evoked responses
Humans
Models, Neurological
Nonlinear diffusion filtering
Nonlinear Dynamics
Pattern Recognition, Automated
Signal Processing, Computer-Assisted
Single-trials
Transcranial Magnetic Stimulation
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In this paper we present a novel application of denoising by means of nonlinear diffusion filters (NDFs). NDFs have been successfully applied for image processing and computer vision areas, particularly in image denoising, smoothing, segmentation, and restoration. We apply two types of NDFs for the denoising of evoked responses in single-trials in a matrix form, the nonlinear isotropic and the anisotropic diffusion filters. We show that by means of NDFs we are able to denoise the evoked potentials resulting in a better extraction of physiologically relevant morphological features over the ongoing experiment. This technique offers the advantage of translation-invariance in comparison to other well-known methods, e.g., wavelet denoising based on maximally decimated filter banks, due to an adaptive diffusion feature. We compare the proposed technique with a wavelet denoising scheme that had been introduced before for evoked responses. It is concluded that NDFs represent a promising and useful approach in the denoising of event related potentials. Novel NDF applications of single-trials of auditory brain responses (ABRs) and the transcranial magnetic stimulation (TMS) evoked electroencephalographic responses denoising are presented in this paper.
ISSN:0165-0270
1872-678X
DOI:10.1016/j.jneumeth.2009.09.017