Using advanced research techniques they developed, four staff members at ORNL evaluated electroencephalogram (EEG) data for changes in a patient before, during and after a seizure. By analyzing the data, which show the electrical impulses that make up brain waves, they found they could detect a seizure eight to 15 minutes before it occurred.
"These results provide a basis for future work that may detect, and perhaps control, some uncontrollable seizures in thousands of epileptic patients," said Lee Hively, one of the ORNL researchers. "One possible form of this technology is a portable beeper device, including a brain wave monitor and the ORNL seizure prediction scheme.
"The device could alert the wearer when an epileptic episode is imminent, allowing the person to stop any dangerous task, such as driving; seek help; or take medication. A more advanced version might add a feature to direct small electronic impulses to stop the seizure before it occurs." Such a device could build on previous research at other laboratories with living sections of a rat's brain. In their research, scientists used chemicals to induce what appeared to be an epileptic seizure, then applied electrical impulses to force the brain back to normal function.
Hively and ORNL colleagues Ned Clapp, Stuart Daw and Bill Lawkins worked with Dr. Michael Eisenstadt, a neurologist at St. Mary's Biomedical Research Center in Knoxville. Eisenstadt provided the EEG data and medical interpretations. ORNL staff members focused on devising methods to study the millions of measurements.
Complicating this task is the fact EEG data contains not only signals associated with brain activity, but also aberrations that accompany actions such as eye blinks, muscle twitches and chewing. All of these obscure the brain wave signal, so researchers had to develop a method that could correct for these aberrations. The researchers' diverse backgrounds (math/statistics, chemistry, physics and nuclear engineering) helped in developing analysis tools to interpret the EEG data.
In explaining the team's success, Hively said, "We used real-world tools that can handle real-world data. Most other analytical techniques can handle only model data."
Normal brain activity includes seemingly random, or chaotic, features, with local brain regions behaving relatively independently. These features show up on an EEG as weak correlations between measurements at different locations of the brain. In a person experiencing an epileptic seizure, however, brain waves at different locations have a "large periodic component and a strong correlation between locations," Hively said. Furthermore, the analysis of chaotic features clearly showed a transition between the non-seizure and seizure states, lasting eight to 15 minutes and ending with a seizure.
ORNL has filed invention disclosures for seizure detection, seizure prediction and removal of low-frequency artifacts from brain wave data. Low-frequency artifacts, which can be likened to background noise and are due to eye blinks, chewing, and muscle twitches, are inherently present in an EEG and can obscure brain wave information. Researchers believe the artifact removal technique could be used as a non-intrusive monitor of worker alertness during extreme stress, possible drug abuse, or fatigue.
The research was sponsored by the Laboratory Directed Research and Development program at ORNL.
ORNL, one of DOE's multiprogram national research and development facilities, is managed by Lockheed Martin Energy Research Corporation.