‘Superpower Glass’ helps kids with autism to read emotions

A pilot study at Stanford University has recently demonstrated that their AI-powered wearable therapy, Superpower Glass, can help to develop social skills in children with autism by identifying facial expressions and ‘gamifying’ social interaction.

We spoke with the study’s senior author, Professor Dennis Wall, to learn more about the technology and its potential.

What is ‘Superpower Glass’ therapy and how does it work?

Superpower Glass, in a sentence, is a wearable therapy intended to both increase emotion recognition, and improve social attention, in children with autism. A Google Glass device analyses facial traced data produced as people socialise, and then sends this information to a mobile app – which then computes the emotion expressed by the person, and returns the emotion to the child via the Google Glass display.

The smartphone application, linked to the Google Glass device, uses machine learning to recognise eight different facial expressions, namely: happiness, sadness, anger, disgust, surprise, fear, contempt, as well as a neutral expression. It functions as the nucleus of the program – acting as a “game console” and enabling various activity modes and games for the child. The app also provides video playback so that parents can review social events, which helps to reinforce social learning.

The machine learning classifier – utilised by the mobile application – has refined its recognition capability through exposure to hundreds of thousands of faces across the eight expressions.

Why is this technology so important?

Today, one in every 59 children is diagnosed with autism. Unfortunately, the supply of licensed practitioners able to deliver appropriate therapy to these children is unable to keep up with demand, which has led to months-long waiting lists for appropriate care. As such, the importance of clinically-validated mobile interventions, able to reduce the burden on both the patients and the system, is rapidly increasing.

Superpower Glass is an example of such a solution, offering a mobile treatment solution that can be accessed by families from their own homes, without requiring the presence of a licensed professional. This system has the potential to overcome several challenges with the current health care standards – empowering parents and children with immediate at-home therapy.

Where did the idea for the therapy originate?

Stanford computer science undergraduate student, Catalin Voss, was experimenting with using Google Glass (when available) as a direct-to-consumer product for facial expression detection. Independently of this, I was working on computer vision artificial intelligence models that detected autism using short home videos.

These concepts had a natural synergy, given: the importance of early detection; the potential of early intervention to ablate autism symptoms; and the need to move solutions into the home, and onto mobile devices. Because of this, Mr Voss approached me, and – together with Postdoctoral Fellow Nick Haber, and advisors Terry Winograd (computer science) and Carl Feinstein (psychiatry) – the idea of Superpower Glass began to take shape.

What are the results of the therapy?

So far, we have completed feasibility and efficacy studies on over 120 children. In these studies, we’ve shown that Superpower Glass is both a comfortable and appropriate therapy option for children, as young as three old, across the entire autism spectrum.

In a separate longitudinal trial – we asked 14 families to use the device at home for several weeks. At the end of the trial, we found that children experienced an average decrease of 7.14 points in autism severity (based the Social Responsiveness Scale), and nearly all parents reported an increase in eye contact. We didn’t include a control group for comparison in the study, making it hard to draw strong conclusions about the intervention effect, but we did complete a randomised controlled trial (RCT) to directly test the therapeutic effects of the system using standard endpoints. We will submit the results of this RCT for peer review and publication soon.

Where do you see technology like this heading in the future?

Increased self-management

Increasingly, there is a need for health care and data management to move into the hands of the primary stakeholder – the individual or patient. By doing this, we can ensure the continuity and improvement of care, health, and wellness. Our Superpower Glass program, for example, has the potential to change the current healthcare landscape by utilising wearable AI – bringing simple, easy-to-implement treatment and prevention solutions to families sooner than the current standard-of-care allows.

Technology like this can offer a much-needed therapy option to families who: have limited access to resources, have difficulty scheduling appointments, and may currently be on waitlists for standard-of-care therapy. I foresee a breadth of mobile technology solutions like this emerging shortly, and so we will start to observe a shift towards more individualised care and precision health.

A new medical paradigm

More broadly, neuropsychiatric disorders are the single greatest cause of disability, caused by non-communicable disease, worldwide; they account for 14% of the global burden. Both diagnostic and therapeutic solutions in brain health are limited, in part, by a paucity of objective and reliable measures – of both behaviour, and of neurological function. This limitation is particularly true for those that are applicable outside the clinic, which are also hindered by the lack of a consistent dialogue between patients and clinicians.

That said, there is a rapid advancement in ubiquitous computing, and in artificial intelligence and computer vision. What these advancements allow for is the deployment of tools to large patient populations that go beyond the simple mainstream ‘wearables’ – delivering diagnostic, scientific, and therapeutic value, and establishing a communication portal between patients and clinicians.

A new paradigm in medicine is rapidly becoming possible. In this new paradigm, ubiquitous computing tools that benefit consumers will become increasingly accessible, which will produce data that can: validate and deploy new models for healthcare, create a fluid patient-to-clinician dialogue, and concomitantly enable a feedback loop of improved science and preventative intervention. Mobile technologies that fit this paradigm are essential for the future of the maintenance of health, and in particular, mental health.