5 EEG Source Imaging and Multimodal Neuroimaging
101
approaches with different properties, which recent research has sought to combine
into singular techniques that are both fast and accurate. In the following sections,
we will provide a brief background for the complimentary modalities that are combined with EEG and then dig further into some of the algorithmic techniques used.
Please note that much of this research is still ongoing and under discussion; many
of the methods presented here are not as established as those in the source localization section, so the discussion will seek to broadly discuss the major developments
within the field. Moreover, many of these methods exist as possible alternatives to
one another with no single dominant approach or overriding theory. This comes at
the cost of some algorithmic depth, though appropriate papers and groups will be
cited for the readers’ reference.
5.3.1 MEG and EEG Combinations
While the fundamentals underlying their signal detection differ greatly, MEG and
EEG both represent the neural activity originated from a common source—the electric current resulting from the activity of a population of neurons. In contrast to the
EEG’s measurement of scalp electrical potentials through the volume conductor,
MEG instead records the magnetic fields generated by the currents (both primary
and volume) associated with sources in the brain (which can again be modelled as
current dipoles). The amplitude of the magnetic field measureable outside of the head
is on the order of a few hundred femtotesla (10
−15 ). This extremely small magnetic
field is detected by MEG sensors known as superconducting quantum interference
devices (SQUIDs). SQUID sensors are highly sensitive to small fluctuations in magnetic field strength, and therefore require a specially shielded recording environment
to attenuate any external magnetic fields. MEG measurement also generally comes
at a much higher cost than its EEG counterpart - a cost that arises due to both the
shielding of the equipment and the permanent helium-cooling required to maintain
the superconductive property of the SQUID sensors. In practice, a MEG recording is
almost always accompanied by simultaneous EEG recording, providing possibilities
for both parallel and integrated data analyses.
Like EEG, MEG signals also represent the superimposition of all primary and volume currents induced by the current sources in the brain. Naturally, source imaging
techniques are also employable and desirable for MEG signals. However, the generation of the forward model for MEG signals is distinctly different from that of EEG
due to the fundamental differences signaling characteristics. First, an important feature of MEG measurement is its insensitivity to sources that are radially-oriented to
the scalp—it is only capable of measuring signals from “tangential” sources. Briefly,
this is due to the fact that magnetic coils placed parallel to the scalp will pick up
magnetic fields that are perpendicular to the coil (since the magnetic flux through the
coil is measured). Radial current sources produce a magnetic field that is parallel to
the sensor because of their orientation, making them invisible to the coils (see [1] for
further discussion of the effect of source orientation). Secondly, the magnetic field is
101
approaches with different properties, which recent research has sought to combine
into singular techniques that are both fast and accurate. In the following sections,
we will provide a brief background for the complimentary modalities that are combined with EEG and then dig further into some of the algorithmic techniques used.
Please note that much of this research is still ongoing and under discussion; many
of the methods presented here are not as established as those in the source localization section, so the discussion will seek to broadly discuss the major developments
within the field. Moreover, many of these methods exist as possible alternatives to
one another with no single dominant approach or overriding theory. This comes at
the cost of some algorithmic depth, though appropriate papers and groups will be
cited for the readers’ reference.
5.3.1 MEG and EEG Combinations
While the fundamentals underlying their signal detection differ greatly, MEG and
EEG both represent the neural activity originated from a common source—the electric current resulting from the activity of a population of neurons. In contrast to the
EEG’s measurement of scalp electrical potentials through the volume conductor,
MEG instead records the magnetic fields generated by the currents (both primary
and volume) associated with sources in the brain (which can again be modelled as
current dipoles). The amplitude of the magnetic field measureable outside of the head
is on the order of a few hundred femtotesla (10
−15 ). This extremely small magnetic
field is detected by MEG sensors known as superconducting quantum interference
devices (SQUIDs). SQUID sensors are highly sensitive to small fluctuations in magnetic field strength, and therefore require a specially shielded recording environment
to attenuate any external magnetic fields. MEG measurement also generally comes
at a much higher cost than its EEG counterpart - a cost that arises due to both the
shielding of the equipment and the permanent helium-cooling required to maintain
the superconductive property of the SQUID sensors. In practice, a MEG recording is
almost always accompanied by simultaneous EEG recording, providing possibilities
for both parallel and integrated data analyses.
Like EEG, MEG signals also represent the superimposition of all primary and volume currents induced by the current sources in the brain. Naturally, source imaging
techniques are also employable and desirable for MEG signals. However, the generation of the forward model for MEG signals is distinctly different from that of EEG
due to the fundamental differences signaling characteristics. First, an important feature of MEG measurement is its insensitivity to sources that are radially-oriented to
the scalp—it is only capable of measuring signals from “tangential” sources. Briefly,
this is due to the fact that magnetic coils placed parallel to the scalp will pick up
magnetic fields that are perpendicular to the coil (since the magnetic flux through the
coil is measured). Radial current sources produce a magnetic field that is parallel to
the sensor because of their orientation, making them invisible to the coils (see [1] for
further discussion of the effect of source orientation). Secondly, the magnetic field is