ARTICLES
Illustrating the Mind
By Prof. Steve Simpson
and the frequency-moment signature method [1] has achieved
results as good as expert visual analysis. The “signatures” are
representations of EEG data collected over 32 seconds, with
each time interval itself divided into 15 smaller windows. For each
window, the strength of the EEG signal is calculated as a function
of frequency, and for the entire signature, central moments of
the distribution of strengths at a given frequency are determined.
Without going into detail, the moments give a measure of how
uniform the signal is at a given frequency: high values at high
moments mean less uniformity.
This article deals with external electroencephalography (scalp
EEG) — the measurement of electrical signals within the brain via
external electrodes on the head, in general, but focussing in on
my own collaborative research. After some background material,
automated analysis of EEG for epileptic seizure detection is
described, followed by an introduction to a new approach for
analysing EEG data.
EEG has been around since the early 1900s, and it is invaluable
for clinical use, in diagnosis and subsequent treatment of
conditions such as stroke and epilepsy. Currently, ambulatory
EEG headsets are becoming available which are far less
cumbersome than the conventional arrangement of thirty or so
electrodes connected by wires to the head. The new headsets
have far fewer electrodes, connect inherently safely via Wi-Fi,
and with proper design can measure signals from the brain while
the wearer is moving around.
The distribution with frequency and with moment is key to
determining whether a particular time interval is representative
of a seizure, and in figure 1A, the two-dimensional frequency –
moment distribution for a typical 32s non-seizure signature is
shown, while figure 1B shows the distribution during a seizure.
There are other parameters, such as electrode location, which
are also important, and further details are given in reference 1.
It doesn’t take much imagination to envisage EEG data collection
and analysis being used in school laboratories in the future.
Artefacts
Artefacts are a serious complication for EEG. These are extraneous
signals that arise from muscles and potentials associated with the
eyes, in particular. They cannot be eliminated, only minimised.
In some applications, the signals from artefacts themselves are
useful. For example, is someone operating dangerous equipment
(such as a car) fully awake? Or for external control of devices by
clenching jaw muscles and so on, and in this context, for assisting
people with disabilities.
It is worth mentioning that some companies marketing EEG
headsets for general use make science-fiction-type claims about
what can be measured and interpreted, but in fact, they are
predominantly measuring artefacts rather than true EEG.
Figure 1 Comparison of a non-seizure EEG signature (A) with a
seizure signature (B). [1]
Epilepsy application
It can be seen that the signatures are very different, and this
difference allows computer software to detect epileptic seizures
successfully.
Epilepsy affects about 1% of the world’s population. It is a chronic
neurological disorder that disrupts normal brain function with the
spontaneous occurrence of seizures, leading to risk of serious
injury and a feeling of helplessness. Diagnosis and treatment are
made more difficult because of the wide range of pathologies
that fall under the epilepsy umbrella, and there are no effective
seizure control therapies for 25% of sufferers.
Frequency-moment signatures provided useful insights into
mental activity in general. Furthermore, the visualisation of the
data can take other forms. With the use of Wi-Fi headsets, the
technology can move from the clinical research environment to
inherently interesting, even inspiring, general applications.
Automated seizure detection (by computer) can aid diagnosis
because visual interpretation of EEG waveforms is not required,
This area is described in the following section.
27
SCIENCE EDUCATIONAL NEWS VOL 67 NO 3