One in 88 children has been diagnosed with an autism spectrum disorder, making the condition more common than childhood cancer, AIDS, diabetes, and spina bifida combined. This creates a public health problem: We’re always going to have more people with ASD than experts to assess and teachers to assist them.

Yet much of today’s research doesn’t have a direct impact on the people who are living with ASD or those who care for them. It’s primarily focused on uncovering what causes the disorder, and we’re a long way from understanding that.

Also, most research focuses on a convenience sample of high-functioning children with ASD who have normal IQs and good verbal ability, the mildest form of the disorder. They are “convenient” because they can go to a lab with an unfamiliar person for some undefined period of time and perform tasks they’ve never done before—all of which requires a lot of self-regulation.

But at least 30 to 50 percent of children on the autism spectrum are too severely impacted to comply with such methodology. These are the kids we understand the least and need to help the most. And these are the kids—when conducting science in standardized, well-controlled settings—who find it difficult to provide data.

So we’re taking the lab to them. I work with computer scientists and electrical engineers to make that happen—experts who create sensors that can be woven into clothes, embedded into accessories, or inserted into devices that can be carried or worn. The devices continuously record physical activity patterns and autonomic nervous system sensing—that is, how a body is responding biologically.

To interpret the data, I also need the context of that information: where the person is and what he is doing. So we also “instrument spaces” with video cameras, microphones, and radio-frequency identification tags.

By bringing these the technologies together, we get powerful information in natural settings—at home, at school, and in the community—about what is happening to an individual with more challenging forms of ASD.

My collaborators and I like to use the analogy of medical imaging to articulate our approach. X-rays, CAT scans, and MRIs are ways to noninvasively look inside the body, identify a troublesome area, measure some property about it, and then target intervention to that area. We’re essentially doing the same to understand communication, socialization, and behavioral development, in what we call “behavioral imaging.”

For example, approximately 75 percent of people with autism engage in repetitive motor movements that are poorly understood—hand flapping, body rocking, and finger flicking. Our wireless and wearable sensors collect information on movement patterns so we can better understand why they’re engaging in those behaviors instead of just telling them to “stop doing that, it looks weird.”

We’re finding in some that physiological arousal is predictive of a child’s performance on a task or a precursor to certain kinds of behaviors. This gives me a better sense of causal relationships between overt behavior and internal physiology, suggesting that the behavior isn’t defiant or pathological, but instead an attempt to self-regulate.

Luckily, when given a choice of selecting an activity, many autistic children gravitate toward computers, DVDs, and dynamic media. This is helpful because we’re developing ways to use a webcam and website running in the background that can automatically recognize facial expressions and record a child’s physiological state (i.e., calculate heart rate) using computer vision algorithms.

These and other emerging technologies can record whether a child is calm or agitated, which is especially useful in predicting the behavior of nonverbal individuals, and allow us to appropriately adjust our interaction styles with them—key information for parents and caregivers.

For instance, take an autistic child who appears to be calm and staring off into space. A teacher might think he’s daydreaming, when in reality, his resting heartbeat is 150 beats per minute. He’s agitated, and if a caregiver were to demand performance rather than, say, help him relax, there might be a bad outcome.

Our goal is to learn as much as we can about what distinguishes children with and without ASD and the most appropriate teaching methods, not only to provide a better life for those affected by ASD, but also to train parents, caregivers, and teachers so there’s less reliance on the limited number of experts.

This is ambitious and will take a long time, but we’re confident based on the successes we’ve had that our computational approach will prove useful to many living with an ASD here and now.

Matthew Goodwin is an assistant professor with joint appointments in the College of Computer and Information Science and the Bouvé College of Health Sciences. He has studied autism for almost 20 years, and co-directs Northeastern’s new doctoral program in Personal Health Informatics.