50
E. Liebenthal and T. Singhal
the last century. The discovery that small fluctuations in electrical potentials can
be measured from the human scalp, a method termed scalp electroencephalography
(EEG), was first reported by the German psychiatrist Berger in 1929. Magnetoencephalography (MEG), or the recording of magnetic fields produced by the electrical
currents in the brain, was first achieved in the late 1960s [21, 22]. Both EEG
and MEG have since evolved to include more recording channels and improved
amplification technology. From a practical standpoint, EEG devices are widely
available, have relatively low cost, and can also be portable and used for longterm recordings. On the other hand, multichannel MEG is more practical than
multichannel EEG because EEG (but not MEG) requires manually establishing
contact between each channel sensor and the scalp. Nevertheless, MEG scanners
are still scant, generally limiting the use of this technology.
EEG and MEG both measure currents arising primarily from excitatory and
inhibitory postsynaptic potentials along the dendritic tree of pyramidal neurons in
the cerebral cortex [88]. Neurons in large areas of cortex, on the order of a few
square centimetres, must be synchronously active to generate a detectable electrical
or magnetic field on the scalp. Both EEG and MEG have exquisite temporal
resolution on the order of 1 millisecond, highly superior to the temporal resolution of
fMRI. Despite the correspondence in the neural origin of the EEG and MEG signals,
there are important differences that result in different sensitivity of the techniques
in certain brain areas. Scalp EEG is sensitive to neural sources generating electrical
fields both tangential and radial to the scalp, whereas MEG is sensitive primarily to
sources generating electrical fields tangential to the scalp. Thus, MEG selectively
measures activity in the grey matter of the sulci of the brain, whereas scalp EEG
measures activity both in the grey matter of the sulci and gyri. Magnetic fields are
also less distorted by the resistive properties of the skull and scalp, and they decay
faster as a function of distance from the source, than electrical fields. Thus, MEG has
increased sensitivity and superior spatial resolution for superficial cortical activity,
compared to EEG [2, 43].
Modelling of EEG and MEG scalp activity to localise the neural source(s) is
fundamentally ill posed because no unique solution exists to this inverse problem
[117]. However, modelling of MEG activity is somewhat simplified by the fact that
fewer sources are identified with this technique and at higher spatial resolution.
Generally good agreement has been reported between EEG and MEG source
localisations, although differences exist and can be attributed to the differences
in sensitivity to source orientation discussed above. Thus, the techniques are best
considered complementary [7, 69].
EEG and MEG responses associated with specific external or internal events,
termed event-related potentials (ERPs), can be obtained by simple averaging of
many epochs, aligned by the time of occurrence of the event of interest [31, 132].
ERP waveforms consist of a sequence of peaks characterised by their polarity,
absolute and inter-peak latency, type of stimuli and experimental conditions that
elicit them, and scalp topography. Latency measures are by far the most useful
aspect of ERPs for clinical applications. Deviations in peak latency measures can
indicate abnormal neural conduction, for example, as a result of demyelination in
E. Liebenthal and T. Singhal
the last century. The discovery that small fluctuations in electrical potentials can
be measured from the human scalp, a method termed scalp electroencephalography
(EEG), was first reported by the German psychiatrist Berger in 1929. Magnetoencephalography (MEG), or the recording of magnetic fields produced by the electrical
currents in the brain, was first achieved in the late 1960s [21, 22]. Both EEG
and MEG have since evolved to include more recording channels and improved
amplification technology. From a practical standpoint, EEG devices are widely
available, have relatively low cost, and can also be portable and used for longterm recordings. On the other hand, multichannel MEG is more practical than
multichannel EEG because EEG (but not MEG) requires manually establishing
contact between each channel sensor and the scalp. Nevertheless, MEG scanners
are still scant, generally limiting the use of this technology.
EEG and MEG both measure currents arising primarily from excitatory and
inhibitory postsynaptic potentials along the dendritic tree of pyramidal neurons in
the cerebral cortex [88]. Neurons in large areas of cortex, on the order of a few
square centimetres, must be synchronously active to generate a detectable electrical
or magnetic field on the scalp. Both EEG and MEG have exquisite temporal
resolution on the order of 1 millisecond, highly superior to the temporal resolution of
fMRI. Despite the correspondence in the neural origin of the EEG and MEG signals,
there are important differences that result in different sensitivity of the techniques
in certain brain areas. Scalp EEG is sensitive to neural sources generating electrical
fields both tangential and radial to the scalp, whereas MEG is sensitive primarily to
sources generating electrical fields tangential to the scalp. Thus, MEG selectively
measures activity in the grey matter of the sulci of the brain, whereas scalp EEG
measures activity both in the grey matter of the sulci and gyri. Magnetic fields are
also less distorted by the resistive properties of the skull and scalp, and they decay
faster as a function of distance from the source, than electrical fields. Thus, MEG has
increased sensitivity and superior spatial resolution for superficial cortical activity,
compared to EEG [2, 43].
Modelling of EEG and MEG scalp activity to localise the neural source(s) is
fundamentally ill posed because no unique solution exists to this inverse problem
[117]. However, modelling of MEG activity is somewhat simplified by the fact that
fewer sources are identified with this technique and at higher spatial resolution.
Generally good agreement has been reported between EEG and MEG source
localisations, although differences exist and can be attributed to the differences
in sensitivity to source orientation discussed above. Thus, the techniques are best
considered complementary [7, 69].
EEG and MEG responses associated with specific external or internal events,
termed event-related potentials (ERPs), can be obtained by simple averaging of
many epochs, aligned by the time of occurrence of the event of interest [31, 132].
ERP waveforms consist of a sequence of peaks characterised by their polarity,
absolute and inter-peak latency, type of stimuli and experimental conditions that
elicit them, and scalp topography. Latency measures are by far the most useful
aspect of ERPs for clinical applications. Deviations in peak latency measures can
indicate abnormal neural conduction, for example, as a result of demyelination in