What would it be like if you could look at your watch and catch your mood forecast for the next few days? If you knew your stress levels were likely to be 40% higher tomorrow, would you make sure to go to bed early and get enough sleep? Would you plan to grab coffee with a friend who always makes you laugh early in the day?
That’s the ultimate dream of Rosalind Picard, the founder and director of the Affective Computing Research Lab at the MIT Media Lab. She studies how to predict stress and anxiety, and her startup Empatica sells wearable sensors that pick up on the physiological signals that come from stress. Empatica’s primary product is the Embrace smartwatch–a stylish wearable containing a medical-grade sensor and machine learning algorithms that were designed for people with epilepsy. The device uses the electrical activity on the skin to track the wearer’s seizures, and last week it became the first smartwatch certified as a medical device by the FDA. In a clinical study, which compared the smart watch’s effectiveness to that of three neurologists, the Embrace detected the seizures of 135 patients over 272 days 100% of the time.
But even as Empatica crosses this milestone, the startup wants to expand its focus to other neurological conditions, like autism and depression–and may one day even be able to help people without medical conditions with their day-to-day stress.
Detecting The “Tsunami”: Epilepsy
Matteo Lai, the company’s cofounder and CEO, explains the company’s challenges in terms of waves–literally. If you’re studying waves, first you have to examine a tsunami so you can understand how smaller, more subtle waves work. Empatica’s proverbial tsunami was epilepsy. Now, it can move on to predicting smaller waves, like depression and autism. In comparison to a seizure, the physiological changes in your body when you’re stressed out are like ripples on a pond.
But while detecting anxiety and stress is much harder than detecting a seizure, the reverse is true for prediction. There’s far more data on stress, and you could use data sources other than the body, like a patient’s phone usage, to look for clues. In August 2017, Picard and other members of her lab published a study that used machine learning to predict the moods of undergraduate students at between 78% to 86% accuracy. But that prediction research isn’t ready yet for the Embrace. Picard sees personalized suggestions for how to improve mood–like seeing a friend, going for a walk, or going to bed early–as an integral part of bringing her research out of the lab and into a consumer smartwatch. So while the predictive technology is improving, Picard wants to hold off on bringing mood forecasts to wearables until she can effectively pair them with a way for you to feel better.
Meanwhile, though the Embrace was recently FDA-certified to detect seizures, predicting them–or another high-stress event, like an autism-related meltdown–is an entirely different challenge, one that Empatica has set its sights on.
“What I’m super excited about with the science [is] that when you get personalized, long-term data from a watch or a phone, we can start to help an individual learn [their] patterns, not on average for some group you may be an outlier in,” Picard says. “We’re not going to be able to 100% say you’ve got a seizure coming of this type at this time. But we can say the conditions are highly likely tomorrow, and you might have a better day.”
Because algorithms can learn about the habits and patterns of a particular person, they could provide truly tailored feedback. This is only possible through machine learning algorithms and huge amounts of data. Lai says that the challenge of predicting seizures is exacerbated by the fact that they’re not typically everyday events–so it will take longer for an algorithm to learn what might cause one person’s seizure versus another. But the potential for helping epilepsy patients have a better understanding of what their personal triggers might be and arm them with the knowledge of when they might be susceptible is monumental.
A Daily Mood Forecast
Lai points out that epilepsy has historically been the key to learning more about how the brain functions. For instance, epilepsy patients have been integral in investigations into how the two hemispheres of the brain function. Similarly, the Embrace’s data on epilepsy might have implications for people with autism and depression, and even people who just get stressed out a lot.
People typically don’t understand or investigate the relationships between different elements in their lives, like sleep and exercise, and how they’re feeling. “Whether it’s pain or stress or depression, typically it’s something people are not good at understanding or communicating,” Picard says. “A lot of us are just not in tune with what we’re doing to ourselves: Maybe I’m a little tired, or irritable, but so is everyone else. It’s the political climate, it’s the weather. It is all those things, everything around you is giving you micro-nudges in a direction, and they’re so small you don’t notice it’s a problem. But if you add up 100,000 of them from the last few months, you’re suddenly in a much worse state.”
By giving people awareness of their patterns, a wearable could empower them to change. Tracking your mood–and not just your self-reported version of your mood, which Picard says is rarely accurate–could, perhaps, be a tool to tackle your stress, in whatever form it takes. That’s where the Embrace comes in–it’s a platform that could track and make sense of all of these things happening in your brain. One beautiful, medical-grade wearable to rule them all.
That is, if people opt-in. Picard is adamant that the company never collects health data or uses it for any research without express consent of Embrace’s wearers. “It is important that we be super absolutely respectful and treat the data with the greatest privacy and sensitivity,” she says. A wearable that collects personal data makes sense for someone suffering from a neurological disorder that can make everyday life difficult. At the same time, in a world where more people are wary of personal data collection, the rest of us will have to decide: Do we want to collect even more personal data about ourselves, if it could potentially make us healthier?