Travis, 19, is eager
to drive, swim, and climb mountains. His doctor, however, has warned
him against undertaking these activities, because they could injure
or kill him. Travis is one of almost 3 million epileptics in America.
During a seizure, his muscles contract violently, and he briefly loses
bladder control and consciousness, often embarrassing and sometimes
hurting himself. His condition has not responded to drug therapy (which
causes many epileptics to be drowsy, uncoordinated, and disoriented),
so Travis risks sudden death from an accident, interrupted breathing,
or heart failure. He avoids social situations and has trouble holding
a job. His medical bills are huge.
ORNL researchers
are developing new technology that could someday improve life for epileptics
like Travis. Indeed, they have devised a computer-based method that
will warn an epileptic that a seizure might occur in the next 20 minutes
or so, giving him time to stop hazardous activities and get medical
help to prevent or reduce the severity of the seizure.
This ORNL work
could eventually lead to a portable, noninvasive monitor, allowing Travis
more freedom. A wearable monitor is greatly preferable to a wall-powered
electroencephalograph (EEG) that records brain activity from a "subdural"
electrode under the skin of his skull. Instead, Travis might wear dime-sized
electrodesone to replace the earring he currently wears and the
other attached by conductive glue to his scalp covered by his baseball
cap. The electrodes would relay EEG signals to a pocket computer that
looks for pattern changes in his brain waves and alerts him that a seizure
is imminent.
![Developers of the SeizAlert method (jpg, 68K)](p21.jpg) |
Ned
Clapp, Jr., Vladimir Protopopescu, and Lee Hively developed the
SeizAlert method to detect the onset of a brain seizure.
|
The seizure alerting
systemdubbed SeizAlertdetects the change from nonseizure
brain waves to patterns that forecast a seizure. SeizAlert was developed
by Lee Hively and Ned Clapp, Jr., both of ORNL's Engineering Technology
Division; Vladimir Protopopescu of the Computer Science and Mathematics
Division; and Paul Gailey, an adjunct staff member in the Energy Division
and a professor at Ohio University. Using DOE funding, the group is
working with Nicolet Biomedical Inc., in Madison, Wisconsin, to develop
a commercial version of the system under a cooperative research and
development agreement.
SeizAlert is a
nonlinear technology that converts continuous time-serial data to a
distribution function for the baseline brain-wave activity. "We then
compare nonseizure activity with pre-seizure and seizure activity,"
Hively explains. "We measure the dissimilarity between the base case
and test case distribution functions to detect pre-seizure conditions."
SeizAlert is sensitive
to many seizure types, according to Hively. Unlike other such systems,
it has a filter that removes confounding signals such as eye blinkssomething
physicians have been trying to do for nearly a century. The ORNL approach
obtains seizure forewarning from a single-channel scalp EEG, rather
than relying on subdural EEG used by other monitoring methods. It has
highly discriminating dissimilarity measures that detect small differences
in a patient's EEG data that are not recognizable by visually scanning
typical EEG charts. SeizAlert provides a warning if these differences
exceed a predetermined threshold.
To fine tune the
algorithm and establish a database for reference, the ORNL researchers
analyzed 19 sets of time-serial EEG data, each from a different patient
who suffered an epileptic seizure. "We saw forewarnings of a seizure
as much as three hours before the event, with a typical forewarning
of 20 to 30 minutes," Hively says. "Also, the algorithm correctly reported
no warnings when tested with EEG data that contained no pre-seizure
or seizure activity."
Once perfected,
the SeizAlert technology would help epileptics like Travis have lots
more fun with much less fear.
Beginning
of Article
Related Web
sites
ORNL's
Engineering Technology Division
Computer Science
and Mathematics Division