Detect epilepsy seizures with fewer false alarms
JOHNS HOPKINS (US) — New brain implant software can more accurately detect imminent epileptic seizures, which should significantly cut false alarms and unneeded electrical jolts.
In early testing using brain wave recordings rather than live patients, the upgraded system significantly cuts the number of unneeded pulses of current that would be sent to the brain, an important improvement, says biomedical engineer Sridevi V. Sarma, who is leading the research.
“If you introduce electric current to the brain too often, we don’t know what the health impacts might be,” she says. “Also, too many false alarms can shorten the life of the battery that powers the device, which must be replaced surgically.”
Medication cannot control seizures in about a third of 50 million epilepsy patients worldwide. One solution is to shoot a short pulse of electricity to the brain to stamp out a seizure just as it begins to erupt. But current brain implants trigger too many false alarms.
“These devices use algorithms—a series of mathematical steps—to figure out when to administer the treatment,” says Sarma, an assistant professor at Johns Hopkins University. “They’re very good at detecting when a seizure is about to happen, but they also produce lots of false positives, sometimes hundreds in one day.”
Her new software was tested on real-time brain activity recordings collected from four patients with drug-resistant epilepsy who had experienced seizures while being monitored.
A study published in the journal Epilepsy & Behavior, reports that the system yielded superior results, including flawless detection of actual seizures and up to 80 percent fewer alarms when a seizure was not occurring.
“We’re making great progress in developing software that is sensitive enough to detect imminent seizures without setting off a large number of false alarms,” Sarma says. Further fine-tuning is under way, using brain recordings from more than 100 epilepsy patients at The Johns Hopkins Hospital, where several epilepsy physicians have joined in the research.
Sarma hopes that within two to four her system will be incorporated into a brain implant that can be tested on people with drug-resistant epilepsy.
“There is growing interest in applying responsive, or closed-loop, therapy for the treatment of epileptic seizures,” says Gregory K. Bergey, professor of neurology at Johns Hopkins and director of its Epilepsy Center.
“Devices to do this have been tested in humans, but for this therapy to be useful for the patient with epilepsy requires early detection of abnormal brain activity that is destined to become a seizure.Detection has to be within seconds of seizure onset, before the seizure spreads to cause disabling symptoms such as alteration of consciousness.
“Developing detection methods that can both provide this early detection and yet not be triggered by brain activity that will not become a clinical seizure has been a real challenge. Dr. Sarma’s group appreciates how important this is. The application of their detection algorithms has produced promising preliminary results that warrant further study of more seizures in more patients.”
In trying to solve the seizure false-alarm problem, Sarma drew on her training in electrical engineering, particularly a discipline called control theory. “We decided to start with the origin of the signal in the brain,” she says.
Sarma’s team compared electrical data from the brains of epilepsy patients before, during, and after seizures and studied how this activity changed over time, particularly when a seizure began.
“We wanted to figure out when would be the optimal time to step in with treatment to stop the seizure,” she says. The team members “trained” their system to look for that moment without setting off false alarms.
The new system for seizure detection with reduced false alarms is protected by a patent obtained through the Johns Hopkins Technology Transfer office. Lead author is Sabato Santaniello, a postdoctoral fellow in Sarma’s lab.
More news from Johns Hopkins University: http://releases.jhu.edu/
You are free to share this article under the Creative Commons Attribution-NoDerivs 3.0 Unported license.