Detecting Seizures and their Autonomic
Impact with a Wristband

by Rosalind W. Picard, Sc.D., FIEEE

We never expected this.

Before the Winter break, a student came to me and asked if he could borrow one of our wristband sensors. The sensor captured the so-called "fight-or-flight response" through electrodermal activity[1] measured from the surface of the skin. He wanted to see what was causing stress for his little brother, who has autism and doesn't speak. I said, "Sure, take two" since back then the wires would often break (we later fixed this). He put one on each wrist of his little brother. He could see the data wirelessly as his little brother went around the house, and I could view it at MIT. What I saw next was completely surprising.

As I sat in my office, reviewing the past week's data, I thought, "This day looks pretty typical" and "Normal variation here." Both wrists transmitted signals that went up with mild autonomic stressors, and down with relaxation. Usually the two sides of the body respond with similar signals, and this was the case in the days I had just viewed. Then I clicked to see the next day's data. My jaw dropped. One wrist showed a peak that was greater than ten times the typical stress response. The other wrist showed no response at all.

One wrist showed a peak that was greater than ten times the typical stress response.

My first thought was "The sensors must be broken." But, both? The one that was too high, or the one that didn't respond? I am an electrical engineer by training and began debugging. Nothing sensor-related explained what I saw. The data from both sensors looked fine right before and after. In fact there was a clear "sleep signature" right after. I have probably looked at more electrodermal data than anybody on the planet and I could not think of anything to explain this.

I pondered it overnight and the next day called the student at home. "Hi, how is your family? Sorry to interrupt your break.... Hey, any idea what happened to your little brother at 4pm on December 26?" He said, "I don't know, I'll check the diary." I appreciated that an MIT student might actually have kept a diary. I silently prayed that this event would have an entry. He came back, "What time did you say?" I replied, "4:00". He paused, "That was 20 minutes before he had a grand mal seizure."

"That was 20 minutes before he had a grand mal seizure."

A giant signal on the wrist before a seizure? How could this be? I called another student's dad, brain surgeon and epilepsy specialist Dr. Joseph Madsen at Children's Hospital Boston (CHB). "Hi Dr. Madsen, my name is Rosalind Picard... Is it possible there could be a huge sympathetic nervous system surge 20 minutes before a seizure?" I didn't want to tell him it was just on one side of the body. After all, we were measuring a component of emotion. How could emotion be on just one side of your body? It didn't make sense.

Joe was very nice, "Probably NOT 20 minutes before the seizure starts in the brain, however maybe it could happen before the outward clinical signs." Then he told me how sometimes patients can have their hair stand on end on only one arm before a seizure. Or, they might have goosebumps on only one side. "On just one side?" I told him about the asymmetry. Joe got very interested, we got IRB approval from MIT and Children's Hospital, and we added our wristbands into an approved study, which simultaneously measured EEG, ECG, video, and now also our data - "EDA" - electrodermal activity. Ming-Zher Poh, a brilliant doctoral student at MIT, designed and built even better sensors for logging the data 24/7 and undertook this novel research study as part of his PhD.

What did we find? 80 patients had successful recordings for the 2-7 days they were monitored. The doctors labeled the EEG's for seizures while blinded to our EDA data, using gold-standard video and EEG. We found that 100% of the most severe seizures, the "generalized tonic clonic (GTC)" a.k.a. "grand mal" had significant EDA responses. Also, 86% of the "complex partial seizures" showed EDA surges of more than 2 standard deviations above the average pre-seizure period. In most cases the seizures were generalized to both sides of the brain, and the wristband responses were on both sides of the body. But there was another very big surprise, even more important.

We looked at what the data on the wrist meant related to the brain waves. All of our patients had epilepsy that was not adequately controlled after trying two or more anti-epileptic drugs. Many of them had a period of time after the seizure ended when all their brain waves (measured on the scalp) went unusually flat. This is called PGES="Post-ictal generalized EEG suppression". The EEG showed the seizure had "ended" but their brain waves, instead of going back to normal, looked almost "dead." Fortunately, nobody died. However, these prolonged periods of flat brain wave activity after seizures are top candidates as biomarkers for sudden unexpected death from epilepsy (SUDEP). SUDEP is the number one cause of death in epilepsy (Institute of Medicine, 2012).

We found that our wristband data was highly correlated with how long the brain waves were suppressed after the seizure.

We found that our wristband data was highly correlated with how long the brain waves were suppressed after the seizure. In other words, the bigger the signal on the wrist, the longer the brain waves went flat after the seizure had supposedly ended. This is not a good situation; it should be detected and the person should not be left alone if it happens.

Usually you need to wear an EEG in order to detect this kind of brain wave suppression. Wearing an EEG is not convenient, not comfortable, and not stylish except maybe at MIT parties. You don't want to wear one 24/7, even if you need what it detect. But, we could now estimate a correlate of the EEG suppression from a tiny comfortable stretchy soft wristband - that could look like an ordinary NBA sweatband on the outside, or "Domo" wristband.

The wristband measure suggests that there is deep brain activation that is causing the EDA and autonomic disruption, even though the cortical activity appears as if it were shut down. The whole brain is not shutting down - just the part of the brain that the EEG is reading near the scalp. Otherwise, our sensor would not fire so strongly.

We also found something exciting about the cardiac functioning - it too was correlated with the wrist signal when there was longer brain wave suppression. The larger the signal on the wrist (the "sympathetic arousal" to the heart and skin), the larger the disruption of the "vagal brake" to the heart (the "parasympathetic branch of the autonomic nervous system"). This is something you do not want to have disrupted, especially when the fight-or-flight system is kicking into overdrive.

All of these findings are published in the top medical journals Neurology and Epilepsia, full citations and links below. If you are a patient, please consider to share our findings with your doctor.

Today there is lots of medical, scientific, and patient interest in getting more of the wristband data, and in providing a way for patients to comfortably learn whether their seizures have huge autonomic surges, and post-seizure brain wave suppression. If they do, then it would be advisable for them to talk to their doctor about how to treat seizures that have this disruption. These may be the most important seizures to issue alerts to watch or help the patient.

Important: As Dr. Madsen said, we don't usually find that the signal on the wristband becomes large far in advance of the brain wave-based seizure starting. The "20 minutes before" did NOT happen for most of our patients when we had precise timing of the sensor along with the EEG. Sometimes the sensor does peak in advance, but usually it peaks at the same time that the EEG registers the seizure. It is possible that the "20 minutes before" that we saw originally involved imprecise timing in our sensor/computer clocks and/or possibly in the diary entry. It is also possible that some people have a big signal in advance of their seizures.

We showed that we can build a significantly more accurate detector of seizures using the wristband data than using "motion-based" event detection

Whether or not we can give advance prediction of a particular seizure's onset, we showed that we can build a significantly more accurate detector of seizures using the wristband data than using "motion-based" event detection. All comfortably wearable seizure detectors on the market today use only motion and are essentially "movement monitors." These don't work for most seizures. We found that the duration of movement was unrelated to the duration of the brain wave suppression. The wristband electrodermal data, however, is related to the duration of brain wave suppression. We want to enable more people to see these data and learn if doing so can help save lives.

Who would have ever expected that we would find so much important information about the brain, while reading "emotion" data from the wrist?

[1] Electrodermal activity is measured in many ways. We use skin conductance. An older term that often applies to these measures is "galvanic skin response (GSR)" but the use of that term is no longer recommended by psychophysiologists. Some people also use the term "electrodermal response (EDR)".


Peer-Reviewed Scientific and Technical Articles