Share this article on:

Neurotech-Epilepsy: Wearable EMG Found to Detect Seizures

Hurley, Dan

doi: 10.1097/01.NT.0000530603.94684.87
Back to Top | Article Outline




A new study demonstrates the feasibility of using a wearable electromyography device to detect tonic-clonic seizures.

A wearable electromyography (EMG) device detected 93.8 percent of generalized tonic-clonic seizures (GTCS) in real time, according to a study published online in Neurology on January 5.

The device produced 0.67 false GTCS false alarms per 24 hours, but nearly all of those occurred during the daytime.

The Neurology paper was among the first to demonstrate its results prospectively, using a pre-specified cut-off for determining that a GTCS is occurring. And at nine seconds, its latency in doing so (from the time of onset as measured by an independent observer) is also among the fastest described so far, the study authors and independent experts noted.

“Most studies on seizure detection have been retrospective,” said the first author of the study, Sándor Beniczky MD, PhD, professor of clinical neurophysiology at the Danish Epilepsy Centre at Aarhus University Hospital in Denmark. “We prospectively validated seizure detection in real time, with a predefined threshold. Just as with drugs, we need robust, reliable studies that prove the efficiency of these devices.”

The new paper comes on the heels of two other studies on wearable EMG for seizure detection, both published in November 2017 in the journal Epilepsia. [See “The Pipeline on Wearable Devices for Seizure Detection.”]

Altogether, they demonstrate the growing feasibility of incorporating wearable seizure-detection devices into clinical practice — both to quantify for physicians how well patients are responding to medication, and to help caregivers respond quickly to ongoing seizures.

Back to Top | Article Outline


The current study included 71 patients who had been referred to long-term video-EEG monitoring on suspicion of GTCS. The surface EMG device, called EDDI and manufactured by IctalCare, was placed on each patient's biceps. Mean recording time per patient was 53.18 hours, with no adverse events reported. Seizure detection by the EMG device was real-time, fully automated, and compared to events identified by trained experts who evaluated the video-EEG recordings and were blinded to device data.

With a sensitivity of 93.8 percent, the device detected 30 of 32 GTCS over a total recording time of 3,735.5 hours. The rate of false alarms was 0.67 per day, but two-thirds of the patients had no false alarms. For 24 of 71 patients (34 percent) who did experience a false alarm, the most common reason was due to physical exercise, which accounted for 68 percent of all false alarms. Only two false alarms occurred during sleep.

Dr. Beniczky said the overall rate of 0.67 false alarms per day “still needs to be improved. It's not such a problem for patients, because it happens almost entirely during the day, so they can stop the alarm when it happens. But it is not good enough if we want a truly automatic device.”

His group has not been able to determine why two-thirds of the patients had zero false alarms while two of the 71 had as many as 10.

“There must be something biologically specific for those who tend to have false alarms,” he said. “We tried to find correlations with the level of physical activity or the number of other electronic devices they had around them. We couldn't find anything.”

Back to Top | Article Outline


“Other groups have looked at recorded seizure data and then created computer algorithms that could recognize those events,” said David Spencer, MD, FAAN, director of the Comprehensive Epilepsy Center and professor of neurology at Oregon Health & Sciences University in Portland. “What this group did here, which I think is important, is they took the next step, testing their algorithm prospectively on new patients. Its sensitivity is about the best I've seen. It can work in real time and early in the seizure, so it checks off a lot of boxes.”

The false positive rate of 0.67 events per day was not especially worrisome to Dr. Spencer, because the vast majority happened during the day. “The biggest concern for my patients is having these major seizures at night, while they're sleeping,” he said. “A number of my patients use a baby monitor in their room, with a family member on the other end. They're counting on that family member hearing something when they have a seizure. It's not very reliable, and there are privacy concerns. So this could be much more effective.”



“All of these devices have positive aspects and flaws,” said Brian Litt, MD, professor of neurology, neurosurgery and bioengineering, and director of the Penn Epilepsy Center at the University of Pennsylvania School of Medicine. “In patients who have classic convulsive seizures and don't create artifacts that generate frequent false-positives, these devices can be useful. There are many types of seizures that fall short of classic convulsions that would be good to detect, however, and these devices are less sensitive to them.”

The results reported in Neurology are “an improvement, but only an incremental one,” said Dr. Litt, who is currently studying the use of an implantable EEG monitoring and therapeutic devices. “EMG devices can be quite useful for some patients, but I believe that EEG remains the gold standard for seizure detection and warning,” he said.

The results reported in Neurology are “an incremental improvement,” said Dr. Litt, who is currently studying the use of an implantable EEG. “I still think EEG is going to remain the gold standard,” he said.

In addition to their use as an alert, the wearable devices can also help neurologists in reviewing the semiology of epileptic events. “We have found that the GTCS waveform in sEMG is distinct from other recorded events,” said José E. Cavazos, MD, PhD, professor of neurology and assistant dean at UT Health in San Antonio. Dr. Cavazos is cofounder of Brain Sentinel and senior author of the Brain Sentinel device study, described in the sidebar, “The Pipeline on Wearable Devices for Seizure Detection.”

“We have been making decisions based on faulty data — on only those seizures that patients tell us about — and there is good data showing that even people with GTCSs miss events,” Dr. Cavazos said. “Finally we have access to another source of data that can help us make better diagnostic and medical decisions.”

Tobias Loddenkemper, MD, associate professor of neurology at Harvard Medical School and director of clinical epilepsy research at Boston Children's Hospital, was a coauthor of the Empatica device study also published in Epilepsia in November. He called the Neurology paper “a great stride forward in the field.”

The study he coauthored was not prospectively designed, Dr. Loddenkemper noted, but was meant to find the best algorithm and cut-off point for detecting GTCS. “The studies had different methods because they had different goals,” he said.

Jon Bidwell, a PhD candidate at the Georgia Tech Everyday Computing Lab, also was a coauthor of the Empatica study.

“EMG is definitely a promising approach for detecting seizures,” he said, “but wearing the device on the bicep may be a difficult sell at first as there are relatively few armband health trackers on the market and people may not be used to seeing them worn for extended periods of time within a social context.”

Back to Top | Article Outline


Dr. Beniczky reported no disclosures related to the study. But since he completed the study of the EDDI device, it has been sold to Brain Sentinel, which has asked him to be a consultant. IctalCare, the manufacturer of the EDDI device, granted the investigators free-of-charge use of the device throughout the study period. Dr. Loddenkemper is part of pending patent applications to detect and predict seizures and to diagnose epilepsy with devices different from the ones used in the Epilesia paper. He also has received sensors from Empatica and Affectiva to perform the reported research. Dr. Litt is a cofounder of a startup company, Blackfynn, which does data integration and analytics.

Back to Top | Article Outline


•. Beniczky S, Conradsen I, Henning O, et al Automated real-time detection of tonic-clonic seizures using a wearable electromyography device Neurology 2018 Epub 2018 Jan 5.
    •. Krauss GL, Ryvlin P. Non-EEG seizure detection is here Neurology 2018; Epub 2018 Jan 5.
      •. Onorati F, Regalia G, Caborni C, et al Multicenter clinical assessment of improved wearable multimodal convulsive seizure detectors Epilepsia 2017;58(11):1870–1879.
        •. Halford JJ, Sperling MR, Nair DR, et al Detection of generalized tonic-clonic seizures using surface electromyographic monitoring Epilepsia 2017; 58(11):1861–1869.
          •. Ulate-Campos A, Coughlin F, Gainza-Lein M, et al Automated seizure detection systems and their effectiveness for each type of seizure http:// Seizure 2016;40:88–101.
            © 2018 American Academy of Neurology