1. What are we measuring with M/EEG (and what are we measuring with)
Gareth Barnes
UCL
SPM Course – May 2011 – London

2. The EEG & MEG instrumentation Neuronal basis of the signal
Outline
A brief history
The EEG & MEG instrumentation
Neuronal basis of the signal
Forward models
Here is the outline of the talk
I will start with a brief history of EEG & MEG as recording techniques
We will then focus on the instrumentation without getting into to much details but just a get the broad picture of the measurement principles and to emphasize the type of activity that we obtain on scalp.
This will be the main part of the talk.
Finally, having in mind the general principles of EEG & MEG signal generation, we will stress the importance of modelling and define forward modelling of EEG/MEG data which will be the explicit ground for most of the inference methods you’ll hear about in the other talks.

3. EEG history
1875: Richard Caton ( ) measured currents inbetween the cortical surface and the skull, in dogs and monkeys
1929: Hans Berger ( ) first EEG in humans (his young son), description of alpha and beta waves
1950s. Grey Walter ( 1910 – 1977). Invention of topographic EEG maps.

4. MEG history 1962: Josephson effect
1968: first (noisy) measure of a magnetic brain signal [Cohen, Science 68]
1970: James Zimmerman invents the ‘Superconducting quantum interference device’ (SQUID)
1972: first (1 sensor) MEG recording based on SQUID [Cohen, Science 1972]
1973: Josephson wins the Nobel Prize in Physics
- And goes on to study paranormal activity…
Brian-David
Josephson
David
Cohen

5. SQUIDS It is an ultrasensitive detector of magnetic flux.
It is made up of a superconducting ring interrupted by one or two Josephson Junctions.
Can measure field changes of the order of 10^-15 (femto) Tesla
(compare to the earth’s field of 10^-4 Tesla)

6. Flux transformers There are different types of sensors
Magnetometers: measure the magnetic flux through a single coil
Gradiometers: when more flux passes through the lower coil (near the head) than the upper get a net change in current flow at the inut coil.
There are different types of MEG sensors.
In all cases, they use coils that transform the magnetic flux through the surface coil into a tiny electric current whose intensity will be instantaneously compensated and hence measured by the SQUID. This is the amount of compensatory current that is related to the magnetic field and yield the MEG signal.
Now with a single coil, we get a magnetometer that measures the local magnetic field directly.
Other systems like the one downstairs from CTF is using gradiometers. Here is a 1st order axial gradiometer.
Axial because the z vertical axis is radially oriented with respect to the scalp. 1st order because it is made of two coils and thus computes the first derivative of the magnetic flux locally by computing the difference between the two opposite currents generated in each coil. One can even couple gradiometers to compute higher order differences and planar instead of axial gradiometers also exist.
Gradiometers are less sensitive to distant sources, meaning that they are both less sensitive to noise from the environment and to deep sources.
This is why some MEG systems like the finish Elekta machine do couple magnetometers and gradiometers.
The type of sensors depend upon the machine you have. However, there are some mathematical tranformations that enable you to represent your data in one way or another, a bit like Laplacian operators transform the EEG data into maps of SCD.

8. The EEG & MEG instrumentation
- 269 °C
SQUIDs
Now, the MEG is a little more complex and it is useful to have a quick look inside.
Here is the real aspect of the MEG recording system and here is a schematic of the inside of the Dewar.
So contrary to EEG, the sensors have fixed position and this the subject’s head location that is fit to the helmet.
Inside, the dewar is full of liquid helium and is thermically isolated from the outside thanks to vacuum space all around it.
Sensors
(Pick up coil)

9. Occurrence in English language texts
EEG
fMRI
MEG
Google Ngram viewer
Thanks to Laurence Hunt and Tim Behrens

10. What do we measure with EEG & MEG ?
From a single neuron to a neuronal assembly/column
A single active neuron is not sufficient. ~100,000 simultaneously active neurons are needed to generate a measurable M/EEG signal.
Pyramidal cells are the main direct neuronal sources of EEG & MEG signals.
Synaptic currents but not action potentials generate EEG/MEG signals
What are the current sources within the brain that we measure with EEG & MEG ?
Given our current knowledge of human brain physiology, signals of interest come from the synchronous activity of groups of neurons that are closed to each other and exhibit a similar pattern of activity. Indeed, a single neuron alone won’t be able to elicit a measurable signal on scalp but a whole population of several thousands who activate synchronously can.
In fact, among cortical neurons, pyramidal cells are believed to be the main sources of the EEG and MEG signals. This is because they are roughly parallel to each other and oriented perpendicular to the cortical surface so that when they activate simultaneously, their contributions sum up into a macroscopic current.
Finally, these currents reflect synaptic current mostly and not action potentials. This is because action potentials generate currents in both opposite directions along the axons so that the sum of their effects is closed to zero.

11. Lateral connectivity
-local + -
Holmgren et al. 2003

12. Magnetic field
MEG pick-up coil
Electrical potential difference (EEG)
scalp
skull
cortex
Volume currents
5-10nAm
Aggregate post-synaptic currents
of ~100,000 pyrammidal neurons

13. What do we measure with EEG & MEG ?
From a single source to the sensor: MEG
MEG
EEG

14. Fig. 14. Return currents for the left thalamic source on a coronal cut through the isotropic model (top row) and the model with 1:10 anisotropic white matter compartment (volume constraint, bottom row): the return current directions are indicated by the texture and the magnitude is color coded.
C.H. Wolters et al. / NeuroImage 30 (2006) 813– 826

15. (conductivity & geometry)
The forward problem
MEG
Lead fields
EEG
Head tissues
(conductivity & geometry)
Dipolar sources

16. Different head models (lead field definitions) for the forward problem
Finite Element
Boundary Element
Multiple Spheres
Single Sphere
Simpler
models

17. Can MEG see gyral sources ?
A perfectly radial source in a spherical conductor
produces no external magnetic field.

18. Can MEG see gyral sources ?
Source depth, rather than orientation, limits the sensitivity of MEG to electrical activity on the cortical surface. There are thin strips (approximately 2mm wide) of very poor resolvability at the crests of gyri, however these strips are abutted by elements with nominal tangential component yet high resolvability due to their proximity to the sensor array.
A quantitative assessment of the sensitivity of whole-head MEG to activity in the adult human cortex. Arjan Hillebrand et al. ,
NeuroImage 2002

19. EEG Auditory Brainstem Response
Wave I/II (<3ms) generated in auditory nerve or at entry to brainstem+ cochlear nucleus
Wave III. Ipsilateral cochlear nucleus / superior olivary complex
Wave IV. Fibres leaving cochlear nucleus and/or superior olivary complex
Wave V. Lateral lemniscus

20. Volume 295, Issue 7654, 9 May 1970, Pages 976-979
IS ALPHA RHYTHM AN ARTEFACT?
O. C. J. Lippold and G. E. K. Novotny
Department of Physiology, University College, London, W.C.1, United Kingdon
Abstract
It is postulated that occipital alpha rhythm in man is not generated
in the occipital cortex, but by tremor of the extraocular muscles.
It is thought that tremor modulates the corneoretinal potential and
this modulation is recorded at the occiput because of the
anatomical organisation of the orbital contents within the skull.

21. Summary
EEG is sensitive to deep (and radial) sources but a very precise head model is required to get an accurate picture of current flow.
MEG is relatively insensitive to deeper sources but forward model is simple.

22. Supp_Motor_Area Parietal_Sup Frontal_Inf_Oper Occipital_Mid
MEG Sensitivity to depth
Cornwell et al. 2008; Riggs et al. 2009
Hung et al. 2010; Cornwell et al. 2007, 2008
Over subjects and voxels
RMS Lead field
Parkonen et al. 2009
Timmerman et al. 2003
Supp_Motor_Area
Parietal_Sup
Frontal_Inf_Oper
Occipital_Mid
Frontal_Med_Orb
Calcarine
Heschl
Insula
Cingulum_Ant
ParaHippocampal
Hippocampus
Putamen
Amygdala
Caudate
Cingulum_Post
Brainstem
Thalamus
STN

23. Sensitivity can be improved by knowing signal of interest
Sqrt(Trials)
sqrt(Noise Bandwidth)
200 Trials, 40Hz BW
400 Trials, 20 Hz BW

25. Lead fields Forward problem MEG EEG forward model Lead fields
Dipolar sources

26. The inverse problem Y = g()+ 
The inverse problem (estimating source activity from sensor data)
is ill-posed. So you have add some prior assumptions
Y = g()+ 
MEG
forward
model
EEG
For example,
can make a good guess
at realistic orientation
(along pyrammidal cell bodies,
perpendicular to cortex)

27. Summary
Measuring signals due to aggregate post-synaptic currents (modeled as dipoles)
Lead fields are the predicted signal produced by a dipole of unit amplitude.
MEG is limited by SNR. Higher SNR= resolution of deeper structures.
EEG is limited by head models. More accurate head models= more accurate reconstruction.

28. Occurrence in English language texts
EEG
fMRI
MEG
Google Ngram viewer
Thanks to Laurence Hunt and Tim Behrens

29. Local Field Potential (LFP) / BOLD
Logothetis 2003

30. For example, gamma frequency seems to relate to amount of GABA.
Note that the huge dimensionality of the data allows you to infer a lot more than source location.. (DCM talks tomorrow)
For example, gamma frequency seems to relate to amount of GABA.
Muthukumaraswamy et al. 2009

31. Karl Friston
Arjan Hillebrand
Will Penny
Marta Garrido
Stefan Kiebel
Jean Daunizeau
James Kilner
Vladimir Litvak
Guillaume Flandin
Rik Henson
Rosalyn Moran
Jérémie Mattout
Christophe Phillips
JM Schoffelen