HBM2006: MEG/EEG Brain mapping course MEG/EEG instrumentation and experiment design Florence, June 11, 2006 Lauri Parkkonen Brain Research Unit Low Temperature Laboratory Helsinki University lauri@neuro.hut.fi
Outline MEG/EEG signals; how are they like...... and how to measure them Combining with anatomical MRI Common artefacts Data pre processing Experimental design Replicability of MEG results
MEG/EEG signals Cellular currents in an active neuron population...... give rise to extracranial electric potentials and magnetic fields EEG = measuring the potential differences on the scalp MEG = measuring the extracranial magnetic fields MEG and EEG are different views of the same neural sources
MEG/EEG signal strength Synaptic input Dipole moment = current distance Q=I d excitatory or inhibitory synapses at apical dendrites or close to the cell body Degree of synchronization Within a cortical patch, ~1% of neurons signalling synchronously with a stimulus produce > 80% of the signal [Hari, 1989] Orientation of the primary current in a perfectly spherical conductor, radial currents do not produce net magnetic field outside of the conductor EEG sees both radial and tangential currents d
MEG/EEG signal strength Depth more attenuation the deeper the primary current no magnetic signal from the center of a conducting sphere Cancellation close by activations with simultaneous, opposing currents decrease the signal B=0 E 0 B 0 E 0
What can we then see with MEG? Almost all of the cortex with fissural activity emphasized Hillebrand & Barnes, 2002 Deep sources (brainstem, thalamus) in special cases Tesche 1996, 1997; Parkkonen 2002
EEG Instrumentation
EEG electrodes Provides a proper contact between skin and amplifier Properties depend on the electrode material. Aim to: Small and stable electrode potential: typ. < 100 mv Low electrode impedance: typ. 1 20 kilo ohms Wide frequency response Non magnetic if to be used with MEG Ag/AgCl (silver / silver chloride) excels in most respects most common may require periodic chloriding
EEG electrode placement Standard system is essential for sensor level data comparison International 10 20 system Originally 21 electrodes, later extended to much larger arrays Electrode locations scale according to head size Standard electrode names Jasper HH, 1958
EEG caps Facilitate application of a large number of scalp electrodes Geodesic nets 32... 256 channels Different sizes Conventional caps Images courtesy of Elekta Neuromag Oy and Electrical Geodesics Inc.
EEG channel types Bipolar (differential) channels a dedicated pair of electrodes (+ and ) for each channel for EOG (electrooculogram; monitoring of eye movements and blinks) and EMG (electromyogram; monitoring of muscular activity) Unipolar (single ended) channels common reference ( ) shared by all unipolar channels for scalp EEG reference at a silent location (mastoid bone, ear lobe) location of the reference electrode substantially affects the EEG pattern!
EEG amplifiers EMG ~500 µv EOG ~200 µv EEG 1 50 µv 30 5 30 150 30 150
E.D. Adrian about EEG With present methods the skull and the scalp are too much in the way, and we need some new physical method to read through them... In these days we may look with some confidence to the physicists to produce such an instrument, for it is just the sort of thing they can do... Nature 1944, 153: 360362.
MEG Instrumentation
MEG signals are very weak 1015 1010 105 MRI magnet Earth's steady magnetic field Magnetic signals from the human heart Muscular activity Alpha rhythm Environmental and instrument noise in a shielded room
The first MEG measurement Weakness of the cerebral magnetic fields: detection practical only with a SQUID sensor First MEG signals (human occipital alpha rhythm) were recorded in 1972 by David Cohen Cohen, 1972
MEG sensor Flux transformer Picks up and squeezes extracranial magnetic flux into the SQUID SQUID (Superconduction Quantum Interference Device) Requires superconductivity and thus low temperatures; immersion in liquid Helium ( 269 Celsius) Outputs a magnetic flux dependent voltage Requires superconductivity Image courtesy of R. Salmelin
MEG system SQUIDs and flux transformers in a cryogenic vessel (Dewar) filled with liquid Helium (4.2 K = 269 Celsius)
Magnetically Shielded Room for MEG A passive shield against environmental magnetic noise Concentric shells of mu metal and aluminum, typically 2 or 3 shells Shielding can be enhanced with active compensation systems Shielding provided by a single shell External magnetic field
MEG sensors: Pick up coil types Geometry of the pick up coil determines its sensitivity pattern most sensitive to sources around the sensor loop most sensitive to sources right beneath the sensor
Auditory data with different sensors Magnetometers, axial gradiometers Planar gradiometers Sensor type essential knowledge for correct interpretation of sensor level MEG data!
MEG: Noise cancellation Reference sensor array a small set of SQUID sensors further away (~30 cm) from the brain to monitor only the external interference weighted output subtracted from the normal channels to reduce external noise [Vrba, 2001] Signal space projection (SSP) signal patterns identified (often statistically) as interference removed from the signal [Uusitalo 1997; Parkkonen, 1998] Signal space separation (SSS) unique decomposition to inside and outside originating fields as dictated by the laws of physics [Taulu, 2005]
Co registration with anatomical MRI/CT MEG source locations usually superimposed on anatomical MR images: MEG/MRI co registration required. Head coordinate frame is the link between the MEG and MRI device coordinate frames. 3 common anatomical landmarks are used to define the head coordinate frame. Note: No consensus among MEG vendors for a common Right preauricular coordinate system! point (RPA) Left preauricular point (LPA) Nasion
Determining head position in MEG 3 to 5 small coils are employed for localizing subject's head in the MEG device coordinate system, i.e., with respect to the sensor array. 3D digitizer is used for localizing the HPI coils in the head coordinate frame prior to the MEG measurement (= digitization)
MEG/MRI integration Coordinate systems MRI device <= landmarks => Head <= coils => MEG device Bridges
Simultaneous EEG and MEG Motivation: MEG and EEG provide complementary information One practical approach: look at the residual EEG when the contribution of MEG identified sources removed Auditory N100 response EEG Clinical applications: MEG + EEG should be used, e.g., in epilepsy diagnostics for optimal sensitivity to epileptic spikes Problems: different bias => difficulties in weighting when combining MEG and EEG for source modelling MEG
Artefacts and Noise Biological muscular / cardiac ocular (blinks, saccades) head movement Moving magnetic material/particles dental work, braces, surgical plates Urban environment power lines (50/60 Hz + harmonics) traffic elevators Wide band noise uninteresting brain activity, brain noise system: SQUIDs and thermal insulation Hari 2000
Pre processing: Removing artefacts 1. Prevent rather than compensate! 2. Discard contaminated periods reject epochs with excessive amplitudes 3. Compensate by removing the artefact field pattern improve the SNR of the artefact pattern by averaging and then remove it by signal space projection eye blinks, cardiac signals 4. Compensate by modelling the source of the artefact (magnetic particle, electric current,...) 5. Use high/low pass filters 6. Head motion artefacts: Track head movement and compensate by SSS Jousmäki 1996
Pre processing: Averaging trial 1 trial 3 trial 2 trial N trial 4 MEG/EEG channels ( + + + + average response )/N= Stimulus trigger channel Signal model: stimulus locked activity + uncorrelated noise Signal recovery: stimulus locked time domain averaging SNR ~ trials Linear operation: Order interchangeable with other linear operations
Pre processing: Filtering Narrow the bandwidth to gain in signal to noise ratio For typical evoked responses: 0.1 40 Hz pass band (except SEFs 0.1 100 Hz) Filters can mislead when used incorrectly Abolished or distorted responses: too narrow pass band, too sharp filters False responses: zero phase shift high pass filters
A MEG/EEG experiment Stimuli (if any) auditory visual somatosensory olfactory pain...? MEG/EEG evoked responses changes in oscillatory activity single trials...? Task attend/ignore detect + react imagine observe/imitate...? Behavioral responses limb/finger movement speech...?
Experimental design MEG/EEG responses mostly reflect transient changes in the sensory input rather than sustained activity as fmri. Stimulus sequences Optimize evoked response SNR given the duration of the measurement noise ~ 1 / trials more habituation with shorter ISIs signal decreases ISI ranges from 25 ms... 30 seconds, typ. 1 2 s. multiple categories/conditions often in parallel rather than in blocks ( fmri) oddball paradigms (frequent standard stimulus + intervening rare deviant)
Experimental design: Timing matters Sloppy stimulus timing (jitter) yields smeared MEG/EEG responses. Physiological jitter produces similar effects. The longer the response latency, the longer and smoother the response. Somatosensory evoked fields Precise timing 20 ms random jitter Latency (ms)
Experimental design: Temporal sampling Bandwidth of interest fhp = DC... 1 Hz flp = 100... 2000 Hz Bulk of cerebral MEG/EEG 0.1... 100 Hz Sampling rate fs > 2 * flp The high frequency component: signals up to 900 Hz MEG/EEG sampling >> fmri sampling, where fs = TR ~ 1... 2 seconds No need to synchronize stimulation with acquisition The traditional N20m response: signals below 300 Hz
Replicability of MEG results Same experiment, same subject, 8 runs within 1.5 years auditory stimuli, 1 khz tone, 50 ms FWHM Hanning window, randomly to left/right ear, ~2 s ISI, 100 accepted trials averaged sound level not controlled rigorously : ( sampling at 600 Hz, 0.1 200 Hz pass band Field map of one run
Replicability of MEG results Source modelling filtering 0.1 40 Hz prior to source modelling 2 dipole model: goodness of fit >95% at the N100m peak overlay on anatomical MRIs Source strengths 10 mm