It looks like a new-age fashion accessory, but the black, chunky bracelet on Rosalind Picard's wrist is actually the latest technology for monitoring seizures. Similar to wearable devices for tracking fitness, this experimental wristband comes packed with sensors that measure heart rate, skin temperature and movement patterns. But unlike other fitness monitors, Picard's biometric bracelet also gauges changes in electrodermal activity, or skin conductance — an indicator of the abnormalities in the nervous system that are triggered during many epileptic seizures.

Picard, an electrical engineer at the Massachusetts Institute of Technology (MIT) Media Lab in Cambridge, developed the idea for her 'Embrace' device with her former PhD student, Ming-Zher Poh. The researchers initially created a prototype bracelet called Q Sensor that tracked only electrodermal activity and movement. Two years ago, they showed that this device could accurately detect 94% of large convulsive seizures experienced by children with epilepsy1. What's more, the Q Sensor produced few incorrect signals — less than one false alarm per day on average — and most of these were triggered by forceful, rhythmic motion, such as shaking dice or vigorously playing with a Nintendo Wii.

Portable devices could help the wearer prepare for a seizure by taking rescue medication or lying down. Credit: IctalCare

About one-third of people with epilepsy continue to experience seizures despite an increasing number of available drug treatments and considerable progress in surgery (see page S12 and page S7). A portable device that warns of an impending attack could help people prepare for a seizure by lying down, say, or taking a rescue medication. Existing devices can only record epileptic events that are already occurring — a major limitation for now. Researchers therefore hope to find measurable biomarkers that would warn the wearer of a seizure before it strikes.

Nonetheless, even a device that falls short of that goal could dramatically alleviate the burden of disease. It could provide real-time information on how well a medication is working, or it could help people determine whether they need to seek medical attention. In one of Picard and Poh's studies, for example, they showed2 that surges in electro-dermal activity detected by their device correlated with a measure of brain activity thought to be linked to the risk of sudden unexpected death in epilepsy, a mysterious complication of the disease that is the most common cause of epilepsy-related deaths. “This autonomic information tells you something that turns out to be really important in terms of what's going on inside the brain,” says Picard. “That was a surprise to us, but it's a pretty cool surprise, and that is where we think the potential is: for alerting people with epilepsy to something that is life-threatening.”

So far, tests of the Q Sensor prototype have taken place in hospital. Picard now hopes that the second-generation Embrace device can be used to detect seizures in the home. To achieve this, she has joined forces with Empatica, a company based in Milan, Italy, that already sells a tool for mobile stress monitoring. According to Matteo Lai, Empatica's chief executive, clinical trials involving Embrace are planned for early 2015.

Sensing seizures

Most commercially available systems for 24-hour seizure monitoring currently rely on motion detection. One such system is manufactured by Smart Monitor, based in San Jose, California, which sells a device called the SmartWatch that uses three-dimensional accelerometers to detect the excessive and repetitive shaking movements that occur during a large seizure. Within seconds, the wristwatch issues alerts by text message or phone call to designated family members, who can then summon help in case of serious injury or loss of consciousness.

SmartWatch and other movement-based detectors have a sensitivity of around 90% for identifying generalized tonic–clonic seizures — big, whole-body convulsions — compared with video electroencephalography (EEG), a diagnostic technique that uses video equipment and electrodes on the scalp to record brain waves and behaviour at the same time. Video EEG is the gold standard but is only available in a hospital setting. Some portable EEG systems have been developed for everyday use, but they are uncomfortable and unsightly — most people with epilepsy say they would not wear scalp electrodes in public to obtain seizure warnings, nor do they want implantable alternatives. To increase the sensitivity while providing a device that is comfortable and stigma-free, some researchers are investigating physiological metrics beyond movement patterns. One such trait is electrodermal activity. Another is electromyography, the electrical activity produced by muscles.

Brain Sentinel, a start-up company in San Antonio, Texas, is developing a prototype device that is about the size of a bar of soap and can be worn around the upper arm to detect the electromyography patterns typical of generalized tonic–clonic seizures. Like the SmartWatch, Brain Sentinel's sensor can alert caregivers that a seizure is occurring. At the 2013 American Epilepsy Society's annual meeting in Washington DC, researchers from the South Texas Comprehensive Epilepsy Center in San Antonio reported that the Brain Sentinel device detected 95% of generalized tonic–clonic seizures. The device had been worn by 33 participants, each of whom wore the sensor for an average of two days, and only one false alarm was reported. A larger clinical trial, involving at least 100 participants at 11 medical centres across the United States, is ongoing.

Compared with detectors that rely on movement patterns, “the much lower false-positive rate [of the electromyography-based device] is definitely an advantage”, says Michael Girouard, president of Brain Sentinel. “A high number of false detections can lead to alarm fatigue,” he explains, as people start ignoring the warnings or stop wearing the device. But with the Brain Sentinel detector, “when the system's alert goes off, you're pretty certain somebody's having a seizure”.

Surface EMG trace (centre) on a bicep showing five minutes of a seizure and 'normal' activity with enlargements of (top) 60 seconds of seizure and (bottom) 60 seconds of 'normal' activity. Credit: IctalCare

IctalCare, based in Hørsholm, Denmark, already sells an electromyography-based device, although it is only available in Denmark. According to Isa Conradsen, the company's clinical research manager, another benefit of using electromyography is that it can detect changes during the tonic phase of a generalized tonic–clonic seizure, when the muscles initially stiffen and people often lose consciousness, whereas accelerometers generally have to wait until the clonic phase, when the muscles begin to spasm and jerk. “Looking at surface electromyography,” Conradsen says, “you can get the alarm much earlier.”

There's more to epilepsy than just generalized tonic–clonic seizures, however. “The problem is that people may die of seizures when there are no convulsions,” says John Duncan, a neurologist at University College London and clinical director of the National Hospital for Neurology and Neurosurgery in London.

In these cases, measuring electrodermal activity could prove particularly helpful. In one study2 using the Q Sensor, Picard and Poh, working with a team that included paediatric neurologist Tobias Loddenkemper of Boston Children's Hospital in Massachusetts, found that skin conductance rose significantly in 86% of complex partial seizures, a type of epileptic attack in which people often stare blankly into space but do not exhibit large convulsions. “The clinical presentation of seizures can vary,” says Loddenkemper, who is now running a larger trial involving the same prototype device. “You need the right sensor for the right epilepsy type.”

Other researchers are working on devices that monitor different physiological changes that occur during seizures. At the Holst Centre in Eindhoven, the Netherlands, for example, scientists are developing a wearable electrocardiogram detector to track different kinds of seizure based on variations in heart rhythm. And at RTI International, a nonprofit organization based in Research Triangle Park, North Carolina, researchers are developing a thin harness that straps around a child's chest and incorporates sensors that measure respiratory function, heart rate and body orientation to detect all generalized seizures and some types of partial seizure.

Eventually, all these signals will probably be combined into a single device. Although writing the algorithm for such a multimodal system might prove tricky, as Jacqueline French, a neurologist at New York University's Comprehensive Epilepsy Center, points out. “The more things you measure that you know change in some people's seizures, the more you're likely to see a characteristic pattern for every person.”

Better trials

We're on the edge of a huge advance in epilepsy science.

Wearable seizure monitors are mostly being developed with the patient in mind. But when they have been clinically validated, these platforms should also prove useful for the biomedical research community. “It could change the way we do epilepsy trials,” says Loddenkemper.

Epilepsy trials today that involve drugs, diets or other treatment interventions typically ask participants at home to keep track of their own seizure experiences in a notebook or electronic diary. But this method depends on people accurately recognizing and documenting their seizures, which is a problem because patients tend to be unaware of about half of all seizures recorded during video-EEG monitoring.

By taking direct physiological measurements instead, wearable sensors could dramatically improve diagnostic accuracy in such studies. “We would get a more objective evaluation of the seizure frequency,” says Sándor Beniczky, a neurophysiologist at the Danish Epilepsy Centre in Dianalund. To test this idea, Brain Sentinel, Empatica, Smart Monitor and others are now running or planning trials designed, at least in part, to compare the accuracy of their devices with the self-reporting of seizures by patients.

Standalone systems for detecting seizures could become obsolete if multinationals such as Apple and Samsung were to roll out smart watches with clinical-grade capabilities such as electrodermal activity or electromyography. At that point, says Lunal Khuon, a biomedical engineer at Villanova University in Pennsylvania, “there will be less hardware development and more software development to use the sensors that are already available”.

“We're on the edge of a huge advance in epilepsy science with better recording of seizures,” says French. “It's fundamental that we understand the symptoms before initiating treatment — and it's something that we're not very good at.”