Detecting seizures before they start

Researchers at MIT and Harvard are preparing to carry out trials of a new device for treating epilepsy. If successful, it would be the first such device to automatically detect and treat seizures, says John Guttag, at MIT's Computer Science and Artificial Intelligence Lab, who developed it with colleague Ali Shoeb and Steven Schachter, a neurologist at Harvard Medical School, in Boston.
Currently, more than two million people in the United States alone have epilepsy. And globally it affects one in every 100 people. While about half of them are able to treat the condition with drug therapies, many others fight a constant battle to find the right drugs to target their condition. And, for many sufferers, such as those whose epilepsy is caused by trauma to the brain, drugs are not an option.

Guttag is working on a technological alternative that involves implanting a pacemaker-like device in the patient's chest. Connected to the device is an electrode that wraps around the vagus nerve, a large nerve that runs down from the brainstem through the neck and into the abdomen. This vagal nerve stimulator (VNS) has two modes, says Guttag. One stimulates the nerve electrically at regular intervals. "There is some evidence that this periodic stimulation has a long-term prophylactic effect," he says. But this is hit or miss.

The other "on-demand" mode, which uses more powerful electrical stimulations, can be activated by the patient when a seizure occurs to try to stop it. Although precisely why this works is unknown, there's plenty of evidence that VNS can actually stop seizures, says Guttag.

But there's a catch. To activate the on-demand mode, patients must swipe a magnetic wrist strap across their chest whenever they feel a seizure coming on, explains Harvard's Schachter. Hence, a patient must be able to sense the early signs of a seizure in enough time to do anything about it.
"In my experiences more than half cannot perceive the onset of the seizures," says Schachter. And of those who do manage to use the magnet, only one in four cases results in a reduction of the seizure's severity, he says. This may be due to a latency effect: any delay could be less effective at reducing the symptoms.

"Part of the problem with VNS is that it's not a closed loop system," says Steven Rothman, a neurologist at Washington University in St. Louis, MO, meaning there's no feedback to the device. He points out that it would be more effective if the system itself, not the patient, could detect the seizure.

A new version of Guttag's VNS device, which will be tested on between 10 and 20 patients over the next few months, attempts to solve this problem by providing the device with feedback from the patient. The patient's brain activity will be monitored using an electroencephalogram (EEG) that is continuously analyzed by a detection program. When a seizure is detected, the device will activate an electromagnet hung over the patient's chest, which, in turn, will activate the implanted VNS device.

Initially, the EEG electrodes will be worn as part of a device that looks like a swimming cap, says Guttag. It wouldn't have to be worn all the time, but could be used, for example, when driving. And the long-term goal is a much less conspicuous object ("We could easily put it under a hair piece"). Eventually, the electrodes could be placed permanently under the scalp, he says. Similarly, the electromagnetic triggering mechanism would be integrated within the implanted VNS device. The mechanics of this proof-of-principle set-up are still crude, says Schachter, but the all-important algorithms are reliable.

In fact, the goal is a detection program good enough to sense a seizure much earlier than a patient could. If so, such a device would not only dramatically reduce the severity of seizures, but also might also prevent them.

Another device for detecting seizures, the Response Neurostimulator, developed by NeuroPace in Mountain View, CA, is also under development and currently undergoing clinical trials. And it also involves trying to detect seizures at an early stage. However, instead of stimulating the vagus nerve, it electrically stimulates the brain directly via electrodes implanted on the surface of or inside the brain.

In theory, NeuroPace's device should have an easier job of detecting the seizure, says Brian Litt, a neurologist and bioengineer at the Hospital of the University of Pennsylvania, in Philadelphia, because such detection electrodes can be placed directly on the brain. In contrast, scalp electrodes tend to pick up much noisier signals, he says.

As far as Guttag is concerned, though, VNS has a clear advantage: "It's less invasive, because we're not actually putting anything in the brain." Indeed, a VNS device that could operate automatically would be welcomed by patients, says Litt. If successful, it could do for epilepsy what pacemakers and implantable defibrillators have done for heart conditions, he says. "At the moment, it's the equivalent of saying ‘when you feel a potentially fatal heart rhythm coming along hit yourself in the chest'."