Single-trial analysis and classification of ERP components - A tutorial

Benjamin Blankertz, Steven Lemm, Matthias Treder, Stefan Haufe, Klaus Muller

Research output: Contribution to journalArticle

551 Citations (Scopus)

Abstract

Analyzing brain states that correspond to event related potentials (ERPs) on a single trial basis is a hard problem due to the high trial-to-trial variability and the unfavorable ratio between signal (ERP) and noise (artifacts and neural background activity). In this tutorial, we provide a comprehensive framework for decoding ERPs, elaborating on linear concepts, namely spatio-temporal patterns and filters as well as linear ERP classification. However, the bottleneck of these techniques is that they require an accurate covariance matrix estimation in high dimensional sensor spaces which is a highly intricate problem. As a remedy, we propose to use shrinkage estimators and show that appropriate regularization of linear discriminant analysis (LDA) by shrinkage yields excellent results for single-trial ERP classification that are far superior to classical LDA classification. Furthermore, we give practical hints on the interpretation of what classifiers learned from the data and demonstrate in particular that the trade-off between goodness-of-fit and model complexity in regularized LDA relates to a morphing between a difference pattern of ERPs and a spatial filter which cancels non task-related brain activity.

Original languageEnglish
Pages (from-to)814-825
Number of pages12
JournalNeuroImage
Volume56
Issue number2
DOIs
Publication statusPublished - 2011 May 15
Externally publishedYes

Fingerprint

Evoked Potentials
Discriminant Analysis
Brain
Artifacts
Noise

Keywords

  • BCI
  • Decoding
  • EEG
  • ERP
  • LDA
  • Machine learning
  • Shrinkage
  • Spatial filter
  • Spatial pattern

ASJC Scopus subject areas

  • Cognitive Neuroscience
  • Neurology

Cite this

Blankertz, B., Lemm, S., Treder, M., Haufe, S., & Muller, K. (2011). Single-trial analysis and classification of ERP components - A tutorial. NeuroImage, 56(2), 814-825. https://doi.org/10.1016/j.neuroimage.2010.06.048

Single-trial analysis and classification of ERP components - A tutorial. / Blankertz, Benjamin; Lemm, Steven; Treder, Matthias; Haufe, Stefan; Muller, Klaus.

In: NeuroImage, Vol. 56, No. 2, 15.05.2011, p. 814-825.

Research output: Contribution to journalArticle

Blankertz, B, Lemm, S, Treder, M, Haufe, S & Muller, K 2011, 'Single-trial analysis and classification of ERP components - A tutorial', NeuroImage, vol. 56, no. 2, pp. 814-825. https://doi.org/10.1016/j.neuroimage.2010.06.048
Blankertz, Benjamin ; Lemm, Steven ; Treder, Matthias ; Haufe, Stefan ; Muller, Klaus. / Single-trial analysis and classification of ERP components - A tutorial. In: NeuroImage. 2011 ; Vol. 56, No. 2. pp. 814-825.
@article{84649ae63773402889c20be8e8e6f1a0,
title = "Single-trial analysis and classification of ERP components - A tutorial",
abstract = "Analyzing brain states that correspond to event related potentials (ERPs) on a single trial basis is a hard problem due to the high trial-to-trial variability and the unfavorable ratio between signal (ERP) and noise (artifacts and neural background activity). In this tutorial, we provide a comprehensive framework for decoding ERPs, elaborating on linear concepts, namely spatio-temporal patterns and filters as well as linear ERP classification. However, the bottleneck of these techniques is that they require an accurate covariance matrix estimation in high dimensional sensor spaces which is a highly intricate problem. As a remedy, we propose to use shrinkage estimators and show that appropriate regularization of linear discriminant analysis (LDA) by shrinkage yields excellent results for single-trial ERP classification that are far superior to classical LDA classification. Furthermore, we give practical hints on the interpretation of what classifiers learned from the data and demonstrate in particular that the trade-off between goodness-of-fit and model complexity in regularized LDA relates to a morphing between a difference pattern of ERPs and a spatial filter which cancels non task-related brain activity.",
keywords = "BCI, Decoding, EEG, ERP, LDA, Machine learning, Shrinkage, Spatial filter, Spatial pattern",
author = "Benjamin Blankertz and Steven Lemm and Matthias Treder and Stefan Haufe and Klaus Muller",
year = "2011",
month = "5",
day = "15",
doi = "10.1016/j.neuroimage.2010.06.048",
language = "English",
volume = "56",
pages = "814--825",
journal = "NeuroImage",
issn = "1053-8119",
publisher = "Academic Press Inc.",
number = "2",

}

TY - JOUR

T1 - Single-trial analysis and classification of ERP components - A tutorial

AU - Blankertz, Benjamin

AU - Lemm, Steven

AU - Treder, Matthias

AU - Haufe, Stefan

AU - Muller, Klaus

PY - 2011/5/15

Y1 - 2011/5/15

N2 - Analyzing brain states that correspond to event related potentials (ERPs) on a single trial basis is a hard problem due to the high trial-to-trial variability and the unfavorable ratio between signal (ERP) and noise (artifacts and neural background activity). In this tutorial, we provide a comprehensive framework for decoding ERPs, elaborating on linear concepts, namely spatio-temporal patterns and filters as well as linear ERP classification. However, the bottleneck of these techniques is that they require an accurate covariance matrix estimation in high dimensional sensor spaces which is a highly intricate problem. As a remedy, we propose to use shrinkage estimators and show that appropriate regularization of linear discriminant analysis (LDA) by shrinkage yields excellent results for single-trial ERP classification that are far superior to classical LDA classification. Furthermore, we give practical hints on the interpretation of what classifiers learned from the data and demonstrate in particular that the trade-off between goodness-of-fit and model complexity in regularized LDA relates to a morphing between a difference pattern of ERPs and a spatial filter which cancels non task-related brain activity.

AB - Analyzing brain states that correspond to event related potentials (ERPs) on a single trial basis is a hard problem due to the high trial-to-trial variability and the unfavorable ratio between signal (ERP) and noise (artifacts and neural background activity). In this tutorial, we provide a comprehensive framework for decoding ERPs, elaborating on linear concepts, namely spatio-temporal patterns and filters as well as linear ERP classification. However, the bottleneck of these techniques is that they require an accurate covariance matrix estimation in high dimensional sensor spaces which is a highly intricate problem. As a remedy, we propose to use shrinkage estimators and show that appropriate regularization of linear discriminant analysis (LDA) by shrinkage yields excellent results for single-trial ERP classification that are far superior to classical LDA classification. Furthermore, we give practical hints on the interpretation of what classifiers learned from the data and demonstrate in particular that the trade-off between goodness-of-fit and model complexity in regularized LDA relates to a morphing between a difference pattern of ERPs and a spatial filter which cancels non task-related brain activity.

KW - BCI

KW - Decoding

KW - EEG

KW - ERP

KW - LDA

KW - Machine learning

KW - Shrinkage

KW - Spatial filter

KW - Spatial pattern

UR - http://www.scopus.com/inward/record.url?scp=79955024787&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79955024787&partnerID=8YFLogxK

U2 - 10.1016/j.neuroimage.2010.06.048

DO - 10.1016/j.neuroimage.2010.06.048

M3 - Article

VL - 56

SP - 814

EP - 825

JO - NeuroImage

JF - NeuroImage

SN - 1053-8119

IS - 2

ER -