MEG videos - Theory
Lauri Parkkonen:
1. Introduction to MEG
2. Principles of MEG and EEG
Welcome to this module on MEG basics! The following lectures will give you the theoretical background you need for MEG practice. You may expect information about the neurophysiological bases of M/EEG signals, advantages of MEG compared to other neuroimaging techniques and how are brain responses recorded in MEG. Besides, most frequent artefacts and basic steps in MEG data analysis are described, along with and a brief overview of the instrumentation needed.
Neurons are said to be active when they “talk to each other”, generating electromagnetic activity. EEG measures this activity through the potential distributed on the scalp, while MEG picks it up from the magnetic fields that extend from them outside the head.
Both MEG and EEG reflect the changes in neural activity over time, by tracking responses to external stimuli or time-dependant performance. This is then often combined with MRI to see the anatomical structure.
3. The neurophysiological origin of MEG signals
MEG and EEG signals are weak and need to be summed to be detectable. Temporal summation occurs in post-synaptic activity, which is slower than the presynaptic action potentials. Spatial summation is possible thanks to dendrites of pyramidal neurons, which are perpendicular to cortical surface. Also, concepts in MEG physics: impressed, primary and return (volume) currents; signal depth, cancellation, amplitude and strength factors, dipole moment, etc.
4. MEG and other neuroimaging methods
5. MEG responses
6. Basic concepts in MEG data analysis
7. MEG instrumentation
MEG is a direct non-invasive measurement of neural activity with high temporal resolution and decent spatial resolution. M/EEG have a very good temporal resolution, of the order of milliseconds, while in fMRI is of the order of seconds. However, spatial resolution much better fMRI, so they are worth combining. Regarding invasiveness: EEG/MEG/fMRI are non-invasive compared to PET/SPECT/FCoG/sEEG.
MEG responses are measured for any kind of stimuli or during resting state. There are evoked responses: synchronised to stimuli where we use trial averaging and control for habituation; and there are induced responses: oscillatory changes not locked to stimulation, where we average instantaneous amplitude instead of trials. The latter accounts for changes in functional connectivity between brain regions.
This is an of the steps in MEG data analysis: signal processing techniques to improve signal-to-noise ratio, source modelling techniques to estimate the location of neural activity (i.e. dipole modelling), visualization of the source superimposed on anatomical information, interferences suppression (external, magnetic interferences, biological artefacts…), filtering signal to the desired frequency band, independent component analysis, trial averaging, etc.
As MEG signals are extremely weak, we need a particular kind of instrumentation to pick them up: a magnetically shielded room is required to avoid external interference, given the relative weakness of our signal and superconductive quantum interference devices (SQUIDS) are key to record the signals. Different sensors are used: magnetometers, axial gradiometers and planar gradiometers. Superconductivity is achieved with liquid helium in which the sensors are embedded, kept at -269º Celsius.
Alexandre Gramfort: The impact of tools and modelling assumptions
Neurotechnology is working to maximize neuroimaging spatial and temporal resolution, but better sensors and better statistical tools are needed to make the best of the data we can have. Here, concepts of M/EEG are presented from a computer vision approach, “take home messages” (THM) are given (sensors’ locations/orientations, SNR, etc.) and analytic strategies are compared. Later, advantages of MNE (minimum norm estimates) software for M/EEG are stated as an alternative to promote sparse solutions with non-stationary sources.
Denis Schwartz: Deep source localization and causality in MEG
Can MEG see deep? This video shows an example and several considerations about it: what is deep source localization based on MEG signals, how can we localize deep sources in the brain, how does it benefit other approaches, which particular experimental paradigm should we use, or what improvements can be done. Secondly, causality is addressed: how to get robust measures of causality on MEG signals, comparison and reliability between-subjects and the strong need for cross-validated studies.
Justin Schneiderman: NeuroSQUID
Why High-Tc SQUIDs? MEG technology is introduced (SQUIDS, helium recycling system, etc.). In particular, the difference High-Tc SQUIDS vs. Low-Tc SQUIDS is stressed: High-Tc are simpler, cheaper, and perhaps better as well due to several reasons described (biomagnetism concepts, technologies comparison, signal-to-noise ratio, etc). The hope for this project is to validate technology in neuroscience studies, starting with the High-Tc SQUIDS based MEG findings from the studies presented.
Mathieu Bourguignon: From MEG signals to functional connectivity
Firstly, functional connectivity is introduced with some illustrative examples of how the brain is organised, according to functional segregation and functional integration, how to define each, and we should study both. Secondly, the topic narrows to cortical activity measures with MEG, reflecting communication between the brain activity and the periphery, so-called cortico-kinematic coherence. Thirdly, cortico-kinematic coherence in the context of movement observation is addressed.
Matti Hamalainen:
1. On-scalp MEG will provide ultimate spatial resolution and sensitivity
2. Infant MEG, why and how?
This video is an introduction to M/EEG in multimodal non-invasive imaging (benefits, forward/inverse problem, interpretation of frequency bands in connectivity, sensitivity to cortical sources, etc.). Then, it summarises the evolution of MEG technology (first real-time magnetoencephalogram, developing alternative sensors, etc.) concluding with future lines: possibilities of new sensors and on-scalp meg (benefits from bringing the sensors next to the head).
First, “why” is explored and different MEG systems are presented. It is a multimodal non-invasive brain imaging technique that has benefits for baby scanning. Second, “how” is addressed: differences in infant vs. adult responses, BabyMEG characteristics, localization errors (anatomical sourcing needs manual intervention), infant M/EEG infant models and some examples. Finally, developing steps towards whole-head BabyMEG, forward models and on scalp meg are described.
Michael Faley: High-Tc SQUIDs for MEG and other prospective applications
Hight-Tc SQUIDS in MEG are introduced along with some good reasons to develop them: reducing the running cost of MEG (improve energy efficiency), get rid of helium, etc. Then, several steps are described: fabrication, testing and microstructural/electron-transport properties of graphoepitaxial high-Tc films, SQUIDS and Josephson junctions on MgO substrates. Concluding, there is the possibility that this makes the sensors cheaper (if widely used, so massively produced) and contributes to wider use of MEG.
Ritta Hari: how does a neuroscientist view signals and noise in MEG recordings
This video highlights MEG advantages, and describes concepts like field strengths, what happens below 1Hz, coregistration, filter settings or effect of interstimuli interval. Also, several problems are described: single responses are prone to artefacts, we need reasonable localization accuracy and source differentiation, automatic methods bring surprises, etc. Last, strategies to improve performance are outlined, along with wishes for the future (i.e. adjustable whole scalp coverage).
Stephen Whitmarsh: Introduction to MEG (vs EEG) instrumentation
This video explains the basis of M/EEG signals, and instrumentation, with a focus in the differences between both techniques. Firstly, some basic concepts (i.e. both techniques measure post-synaptic potentials). Secondly, instrumentation is detailed: passive/active shielding, how environmental noise is supressed with an axial gradiometer,etc. Finally, EEG and MEG are compared in detail (i.e. which currents are visible in MEG but not in EEG, or vv.).