The analysis of event-related signals (ERS’s) in neurophysiological studies aims at exploring the information processing in the human brain. When presenting auditory, visual and other stimuli the electromagnetic brain signals recorded as electroencephalogram (EEG) and/or magnetoencephalogram (MEG) (working group 8.21) reflect the corresponding brain activity.
ERS’s are embedded in the spontaneous EEG/MEG activity and background noise and they are typically small in amplitude. Usually, stimulus synchronous averaging is carried out to improve the signal-to-noise ratio (SNR). However, such a procedure does not account for the variability between single ERS’s. In order to avoid this loss of information analysis of single-trial ERS’s has to be carried out. The challenging task for such an analysis is the low SNR together with the fact that single ERS’s and spontaneous EEG/MEG activity have large spectral overlap.
Signal processing procedures, currently developed at PTB, focus on the estimation of the single-trial ERS parameters amplitude and latency. By means of suitable bandpass filtering and application of the Hilbert transform the relevant spectral contents of an ERS can be decomposed into two independent signals, envelope and phase. From these signals then amplitude and latency can be derived.
The bandpass filtering procedure is illustrated using averaged auditory event-related fields (MEG) to sounds of 125 Hz and 1000 Hz. Figure 1 shows the averaged signals and the derived envelope and sine-phase signals. Within the time interval from 100 ms to 150 ms the sine-phase waves can be used to determine latency differences between the event-related fields to sounds of 125 Hz and 1000 Hz.
For EEG/MEG recordings typically an array of spatially distributed sensors is used. This enables the construction of a spatial filter, which can substantially improve the estimation of parameters from single-trial ERS’s. Spatial filtering aims at the suppression of signals from interfering sources, e.g. the spontaneous activity, while leaving the ERS’s unaffected.
The effect of the different filtering steps upon single-trial ERS’s is shown in Figure 2 for one exemplary MEG channel. The spatial filter was constructed from the 93-channel MEG using Noise Adjusted Principal Component Analysis (NAPCA). By spatial filtering a substantial reduction of interfering signal components is achieved and single-trial responses can easily be recognized.
Publications
Publication single view
Article
Title: | Shifted factor analysis for the separation of evoked dependent MEG signals |
---|---|
Author(s): | F. Kohl, G. Wübbeler, D. Kolossa, M. Bär, R. Orglmeister and C. Elster |
Journal: | Physics in medicine and biology |
Year: | 2010 |
Volume: | 55 |
Issue: | 15 |
Pages: | 4219--30 |
IOP Publishing | |
DOI: | 10.1088/0031-9155/55/15/002 |
ISSN: | 1361-6560 |
Web URL: | http://iopscience.iop.org/article/10.1088/0031-9155/55/15/002 |
Keywords: | Evoked Potentials,Humans,Magnetoencephalography,Magnetoencephalography: methods,Models, Statistical,Neurons,Neurons: cytology,SingleTrial |
Tags: | 8.42, Gehirn, SingleTrial |
Abstract: | Decomposition of evoked magnetoencephalography (MEG) data into their underlying neuronal signals is an important step in the interpretation of these measurements. Often, independent component analysis (ICA) is employed for this purpose. However, ICA can fail as for evoked MEG data the neuronal signals may not be statistically independent. We therefore consider an alternative approach based on the recently proposed shifted factor analysis model, which does not assume statistical independence of the neuronal signals. We suggest the application of this model in the time domain and present an estimation procedure based on a Taylor series expansion. We show in terms of synthetic evoked MEG data that the proposed procedure can successfully separate evoked dependent neuronal signals while standard ICA fails. Latency estimation of neuronal signals is an inherent part of the proposed procedure and we demonstrate that resulting latency estimates are superior to those obtained by a maximum likelihood method. |