Navigation and service

Real-time data analysis

MEG is an essential tool with which to study transient changes in electrophysiological processes and is therefore a critical tool for decoding the spatio-temporal organisation of the human brain in vivo. The development of real-time MEG data processing, combined with neurofeedback interfaces, is essential for decoding the human brain as a whole. Such technology would allow the study of the acting and responding brain interactively and would therefore offer a variety of opportunities ranging from basic research to clinical applications. However, the analysis of the acquired data would be challenging and require novel strategies to cope with data processing millisecond by millisecond (Rongen et al., 2006). Once these challenges have been addressed, it is envisaged that the ability to control for fast and transient brain responses, utilising real-time analysis, will offer novel strategies in the fields of neuroscience, diagnosis and therapy.
Standard MEG data analysis usually involves noise and artefact rejection, source localisation and statistical analysis on the spatio-temporal dynamics of the neuromagnetic responses. These analytical steps are computationally demanding and it usually takes in the region of several days to evaluate each experiment. Real-time data analysis on the other hand would offer results during the measurement process and therefore offers great potential in several fields of neuroscience and therapy. Our real-time project relates to the development of real-time data analysis procedures for MEG. The major focus of this project is to establish new concepts of data analysis, which will provide at least first glance results during running MEG measurements.
Recently, we have introduced new methods for noise and artefact rejections that are sufficiently capable for use in real-time applications (Breuer et al., 2014a, b). The major challenge here was handling unaveraged, short data segments during acquisition. The new algorithm is designed for applications using modern, multi-channel MEG systems and can therefore cope with over several hundred MEG channels . Real-time artefact rejection was performed utilising the concept of constrained ICA (cICA). In cICA prior knowledge of an underlying expected signal (e.g., the cardiac activity) is used to optimise the internal cost-function of ICA. Identification of the isolated artefact signals is performed automatically and is included in the artefact rejection scheme. By combining both strategies, it is possible to expedite the whole artefact rejection process.


Figure 1: Source reconstruction of ten trials of auditory evoked fields before (a) and after (b) artefact rejection averaged with respect to the stimulus onset. The trials contain strong cardiac and ocular activity at the time of the auditory responses. On both a and b the source activity is shown for the slice of maximum activity after artefact rejection at the time of the N100m peak.


Chen, T., Suslov, S., Schiek, M., Shah, N.J., van Waasen, S., Dammers, J., 2017. Model-Driven Development Methodology Applied To Real-Time MEG Signal Preprocessing System Design.

Breuer, L., Dammers, J.J., Roberts, T.P.L., Shah, N.J., 2014. A Constrained ICA Approach for Real-Time Cardiac Artifact Rejection in Magnetoencephalography. IEEE Trans. Biomed. Eng. 61, 405–414. doi:10.1109/TBME.2013.2280143

Florin, E., Bock, E., Baillet, S., 2014. Targeted reinforcement of neural oscillatory activity with real-time neuroimaging feedback. Neuroimage 88, 54–60. doi:10.1016/j.neuroimage.2013.10.028

Breuer, L., Axer, M., Dammers, J., 2013. A new constrained ICA approach for optimal signal decomposition in polarized light imaging. J. Neurosci. Methods 220, 30–38. doi:10.1016/j.jneumeth.2013.08.022

Rongen, H., Hadamschek, V., Schiek, M., 2006. Real time data acquisition and online signal processing for magnetoencephalography. IEEE Trans. Nucl. Sci. 53, 704–708. doi:10.1109/TNS.2006.87480