Permanent URL to this publication: http://dx.doi.org/10.5167/uzh-26853
Mandelkow, H. Combining EEG and MRI for the purpose of imaging neuronal currents in humans. 2009, ETH Zuerich.
PDF (currently abstract only. Soon full diss)
View at publisher
Ionic currents generated by neurons in the human brain are arguably the very essence of brain activity and potentially a key to understanding brain function. Present‐day techniques for measuring brain activity non‐invasively in humans include the electroencephalogram (EEG) and the magnetoencephalogram (MEG), both of which have a spatio‐temporal resolution on the order of centimetres and milliseconds. Alternatively, functional magnetic resonance imaging (fMRI), based on the blood oxygenation level dependent (BOLD) effect, offers a resolution on the order of millimetres and seconds. However, none of these techniques applicable to humans offer a resolution on the length and time scale of neuronal processes (roughly micrometres and milliseconds, respectively).
The growing popularity of functional brain mapping in neuroscience creates a strong incentive to further develop non‐invasive imaging techniques with improved resolution. The present work investigates the combination of EEG and MRI measurements and specifically the prospects of directly detecting neuronal currents by MRI, which theory predicts to be possible. This particular EEG‐MRI approach has received growing attention in recent years not only because it promises to unite the high temporal resolution of the EEG with the superior spatial resolution of MRI, but also because such a neuronal current MRI method, if feasible, would reflect electrical brain activity more directly than the detection of associated changes in blood oxygenation, flow and volume by BOLD fMRI.
Simultaneous EEG and MRI experiments are challenging primarily because the MRI scanner generates electromagnetic fields, which strongly interfere with the sensitive measurement of electrical scalp potentials by EEG. Post‐processing algorithms based on average artefact subtraction (AAS) have proven to be instrumental and efficient in removing the notorious MRI gradient artefact (MGA). However, the residual MGA after AAS typically limits the quality and usable bandwidth of the EEG data despite further reduction through re‐sampling, principal component analysis (PCA), and regressive filtering. This work demonstrates the use of a frequency divider and phase‐locking device for the purpose of synchronizing an MRI acquisition with a simultaneous recording of the EEG. Synchronization greatly improves the effectiveness of MRI artefact removal from the EEG signal by AAS. It complements or replaces other post‐processing techniques, thereby increasing the usable bandwidth of the EEG signal to cover the full range of human Gamma band activity up to 150Hz.
In an effort to find an optimal recording and post‐processing strategy for EEG‐fMRI experiments the hardware synchronisation method was compared to and combined with the aforementioned artefact reduction methods based on post‐processing the EEG signal. Comparisons were based on data recorded in vivo and in vitro as well as simulations of the MRI gradient artefact. The simulations developed for this purpose include a framework for quantifying the performance of post‐processing algorithms for EEG‐MRI data. The results suggest a number of improvements to existing EEG‐MRI methodology, but above all they lead to the development of a new software synchronisation method, which substitutes the technically more demanding hardware synchronisation under general conditions.
It has been suggested recently that the influence of the neuro‐magnetic field should make electrical brain activity directly detectable by MRI. To test this hypothesis, we performed combined EEG‐MRI experiments, which aim to localize the neuronal current sources of alpha waves (8‐12Hz), one of the most prominent EEG phenomena in humans. A detailed analysis of cross‐spectral coherence between simultaneously recorded EEG and MRI time series revealed no sign of alpha waves. Instead the EEG‐MRI approach was found to be hampered by artefacts due to cardiac pulsation, which extend into the frequency band of alpha waves. Separate brain displacement mapping experiments confirmed that not only the EEG but also the MRI signal is confounded by harmonics of the cardiac frequency even at 10Hz and beyond. This well‐known ballistocardiogram artefact in the EEG cannot be avoided or eliminated entirely by available signal processing techniques. Therefore we conclude that current EEG‐MRI methodology based on correlation analysis lacks not only the sensitivity but also the specificity required for the reliable detection of alpha waves.
64 downloads since deposited on 20 Jan 2010
15 downloads since 12 months
|Referees:||Boesiger P, Brandeis D|
|Communities & Collections:||04 Faculty of Medicine > Institute of Biomedical Engineering|
610 Medicine & health
|Deposited On:||20 Jan 2010 15:54|
|Last Modified:||09 Jul 2012 04:06|
|Free access at:||Official URL. An embargo period may apply.|
|Other Identification Number:||10.3929/ethz-a-005884782|
Users (please log in): suggest update or correction for this item
Repository Staff Only: item control page