Previous studies indicate that early access to care is vital for children with autism, as beginning therapy during sensitive periods of development—before age 4—leads to greater outcomes and improved social development when compared with care starting later in life. However, the clinical guideline-recommended therapy (20 hours per week for at least 2 years) often has waitlists up to 18 months, as certified clinicians are scarce.
“These expensive, time-consuming therapies are out of reach to many families,” says Dennis P. Wall, PhD. “New, alternative therapies for children with autism need to be developed, especially technology-based therapies, which can overcome many barriers to access that traditional clinician-dependent therapies struggle to overcome.”
For a study published in Nature Digitlal Medicine, Dr. Wall and colleagues developed a machine learning-assisted software system called Superpower Glass (which runs on Google Glass and an Android smartphone) designed for use during social interactions (Figure). Since children with autism spectrum often disorder (ASD) struggle to recognize facial expressions, make eye contact, and engage in social settings, this new technology pairs with a face-identifying artificial intelligence (AI) app that tells wearers which of 8 emotions they’re seeing: happiness, sadness, anger, disgust, surprise, fear, neutral, or contempt.
The Android app that accompanies the Google Glass hardware has three modes:
- “Free play”: children wear the glasses while interacting in their normal environments, capturing faces and emotions wherever they appear.
- “Guess my emotion”: a parent acts out an expression
- “Capture the smile”: children give another person clues about an emotion they’re thinking of until the other person successfully acts it out.
“Before testing the Superpower Glass system in a clinical trial, the first necessary steps were to 1) test the feasibility of use and form factor for children with autism, and 2) identify the potential therapeutic effect(s) and dosage,” says Dr. Wall. “We decided that the best way to test the system was to send it home with families—which is where it would be used most often if it became a therapy—and have the parents and children provide us with feedback about their experiences. Parents also completed behavioral questionnaires at the start and end of the study to see if there were any improvements on the included measures after using Superpower Glass.”
Families who complied with the minimum usage requirements had the device at home for an average of 72 days. In total, there were 5,726 minutes of app usage logged data. Participant mean app usage was 409 minutes, and change in outcome was assessed using the Social Responsiveness Scale (SRS-2).
Of the participating families, 21.4% used the device for 1 month for an average of 10.27 sessions per week, 14.2% used the device for more than 1 month but less than 2 months for an average of 9.14 sessions per week, 33.3% families used it for more than 2 months but less than 3 months for an average of 4.39 sessions per week, 20% used it for more than 3 months but less than 4 months for an average of 3.29 sessions per week, and 7.1% used it for 4.5 months for an average of 3.76 sessions per week.
Among participating parents, 85.7% said their children made more eye contact at the end of the study when compared with baseline. The children’s average scores on parent-completed questionnaires of their social skills decreased by 7.38 points during the study, meaning some autism symptoms had decreased. Nearly half (42.8%) of participants had large enough declines in their scores to move down a step in severity in their autism classification, from “severe” to “moderate” or “mild” to “normal” (Table).
In addition, SRS-2 data showed noteworthy changes between pre-Glass and post-Glass usage on sub-domain questions, including changes in recognizing intent, social initiation, social interaction, and eye contact. The Wilcoxon Rank Sum test run on the SRS-2 65 item-level questions showed a nominally significant change from intake to conclusion.
Dr. Wall says there is a need for healthcare and data management to move into the hands of the primary stakeholder—the individual or patient. “By doing this, we can begin to ensure the continuity and improvement of care, health, and wellness,” he adds. “The Superpower Glass program, for example, has the potential to change the current healthcare landscape by utilizing wearable AI, bringing simple, easy-to-implement treatment and prevention solutions to families sooner than the current standard-of-care allows.”
Google has gifted 50 second addition glasses to Dr. Wall’s lab for further clinical testing, which is underway now. Interested and potentially eligible participants can visit autismglass.stanford.edu to learn more and signup. In addition, Dr. Wall and colleagues have launched a complementary program called “GuessWhat,” which is an enhancement of the Superpower Glass application to enable a form of social therapy with an app on Android and iOS. This system is available at guesswhat.stanford.edu and has no restrictions for eligibility or use at this stage. Thus, it is free and available to all immediately.
Check out Dr. Wall and colleagues’ recent paper highlighting AI on video for rapid, mobile autism detection and measurement at https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1002705.
Daniels J, Schwartz J, Voss C, et al. Exploratory study examining the at-home feasibility of a wearable tool for social-affective learning in children with autism. Nature. 2018, August 2. [Epub ahead of print]. Available at: https://www.nature.com/articles/s41746-018-0035-3.