Wikiversity:Fellow-Programm Freies Wissen/Einreichungen/Unfold: An opensource toolbox for overlapping and non-linear EEG data analysis

Aus Wikiversity
Zur Navigation springen Zur Suche springen

Unfold: An opensource toolbox for overlapping and non-linear EEG data analysis[Bearbeiten]


Background / Motivation[Bearbeiten]

The most dominant method to record brain activity in humans is Electroencephalography (EEG). Due to the bad signal-to-noise ratio, researchers often need to calculate event-related brain potentials (ERPs), that is, an average over many trials. Traditionally, these ERPs are recorded in simple paradigms and well controlled laboratory settings which are strictly controlled to exclude any overlap in brain activity between events. In recent years, however, there has been an increasing interest to expand ERP methodology towards more naturalistic situations that often involve fast or multimodal stimulation, quasi-experimental designs, or complex motor behavior. Unfortunately, in such situations overlapping responses between subsequent events are unavoidable. In addition, brain responses are often influenced by various (low-level) continuous covariates, such as stimulus properties like luminance, contrast or position. These influences are commonly non-linear, for example logarithmic or quadratic in nature. Examples of paradigms where systematic temporal overlap variations and low-level confounds between conditions occur include mobile brain/body imaging (Ehinger 2014) or combined EEG/eye-tracking paradigms (Dimigen 2011, Ehinger 2015). But even “traditional”, highly controlled ERP datasets often contain a mix of overlapping activity (e.g. from stimulus onsets and motor responses), that can be advantageous to disentangle.

Our new open source toolbox "unfold"[Bearbeiten]

In this project we try to develop a powerful, yet easy-to-use open source MATLAB-toolbox for regression-based EEG analyses that combines existing concepts of massive univariate modeling, linear deconvolution modeling, and non-linear spline regression (i.e. generalized additive modeling) into a coherent analysis framework. The toolbox is designed to be modular and can handle even large datasets efficiently. It also includes advanced options for regularization and the use of time basis functions (e.g. Fourier sets). The toolbox allows to estimate reliable neural responses in areas of research confronted with overlapping activity and quasi-experimental designs, such as multimodal ERP studies, combined EEG/eye-tracking experiments, or mobile brain/body imaging research. It can also be applied to other overlapping physiological signals, such as pupillary or electrodermal responses. The toolbox will be freely available at under a GNU open source license.

Current Status[Bearbeiten]

Some of the methods to solve these problems already exist, for example methods for overlap correction, but there is no analysis toolbox available. Other methods, like the non-linear effect estimation, have been previously used, but only in a very limited scope. The main reason is, that there is no toolbox available that allows for easy and flexible application of these analyses. I’m convinced that research is shaped by the tools that are available to the researchers, thus such a toolbox will not only allow analysis of previously impossible to analyze datasets, but will also drive methodological research in itself. I am the main developer of the unfold toolbox developed in matlab. This toolbox can be used to analyze overlapping signals and estimate non-linear effects, simultanteously. In this project I collaborate with Olaf Dimigen (HU Berlin). We use github to plan, debug and document the toolbox. An additional sphinx-based documentation webpage is planned under In order to test the toolbox, we employ unit-tests of all major functions. Parts of the toolbox are already implemented and the technical documentation is partially written. We plan to release the toolbox under an open source license and two accompanying papers in open access journals. Because we are both early career researchers, we currently do not have any funding for this project.

Selected References[Bearbeiten]

Dandekar S, Privitera C, Carney T, Klein SA (2012b) Neural saccadic response estimation during natural viewing. J Neurophysiol 107:1776–1790

Dimigen O, Sommer W, Hohlfeld A, Jacobs AM, Kliegl R (2011) Coregistration of eye movements and EEG in natural reading: Analyses and review. J Exp Psychol Gen 140:552–572.

Ehinger B V., Fischer P, Gert AL, Kaufhold L, Weber F, Pipa G, König P (2014) Kinesthetic and vestibular information modulate alpha activity during spatial navigation: a mobile EEG study. Front Hum Neurosci 8

Kristensen E, Rivet B, Guérin-Dugué A (2017b) Estimation of overlapped Eye Fixation Related Potentials: The General Linear Model, a more flexible framework than the ADJAR algorithm. J Eye Mov Res 10:1–27

Smith NJ, Kutas M (2015b) Regression-based estimation of ERP waveforms: I. The rERP framework. Psychophysiology 52:157–168.


  • Name: Benedikt Ehinger
  • Institution: Institut für Kognitionswissenschaften, Universität Osnabrück
  • Kontakt:
  • Website: