A Device to Spot Autism Early
Researchers have developed a device that can automatically identify autistic children as young as 24 months using the vocalizations they make during a normal day at home. Instead of waiting months or years for an appointment with a specialist, parents could get an objective diagnosis by mail in a couple of weeks.
According to a recent study from researchers at Washington University in St. Louis, the average age of diagnosis for autism is 5.7 years old – several years after such a diagnosis is first possible. With the new system, developed at the LENA Foundation, diagnosis takes just a few weeks. “Intervention is most effective when the child is two to four years old,” says Jill Gilkerson, director of child language research at the LENA Foundation.
The foundation says its tool can distinguish, with 91 percent accuracy, between a child developing normally, a child with autism, and a child with unassociated language delays.
The home kit, called LENABaby, hit the market earlier this month and consists of a questionnaire regarding the child’s development, a digital audio recorder, and an outfit for the child to wear. First thing in the morning, the parent puts the outfit on the child and slides the recorder into a pocket on the front. The recorder is left on all day so that it can capture up to 16 hours of audio. At the end of the day, the parent removes the device from the pocket and sends it to the foundation, where the LENABaby software can analyze the data.
“Roughly speaking, autistic children vocalize differently from other children,” explainsDongxin Xu,manager of software and language engineering at the LENA Foundation. While this concept isn’t new, clinicians have had a hard time using it to their best advantage when making a diagnosis. The problem is, in part, logistical: with existing methods, it is hard to collect enough good-quality data.
Gilkersonsays that most traditional assessments take less than four hours. LENABaby, in contrast, considers a full day of activities in the child’s natural environment. Traditional assessments can be done in the child’s home, but this often involves multiple video cameras and lights, which can influence the child’s behavior.
Even if enough data is gathered, analyzing the audio is still extremely difficult and time-consuming. Making a diagnosis isn’t as simple as counting the number of times the child makes a certain kind of sound, explains Jeffrey Richards, a statistician and database technician for the LENA Foundation.
Richards says the LENABaby software, which he helped develop, starts by breaking down the 16-hour audio stream into segments. Each segment is automatically classified according to the type of sound contained in the clip, such as sounds from the child, a parent, or television. Vocalizations from the child are then assessed further using complex algorithms that look at a variety of factors, such as the phonological composition of the each sound and how sounds are clustered and paired. “We’re simultaneously looking across many dimensions at the same time,” says Richards. Using LENA’s database of previously analyzed audio, the software considers how these characteristics compare to those of children developing normally, children with delayed language development, and autistic children.
While the level of accuracy achieved by LENABaby is high, researchers at the foundation are investigating ways of making the tool even better at diagnosing children. The team is also adapting the tool for use with younger children.
LENABaby can be used for more than a basic diagnosis, helping to track a child’s language development. This could make it a valuable tool for clinicians who otherwise have to rely on data collected during brief, infrequent visits.
LENABaby wasn’t designed to replace clinicians entirely, Gilkerson says, but it does provide “a naturalistic observational component” to an assessment. “Up until now, clinicians had no idea what was going on in the home environment,” Gilkerson says. Analyzing verbal interactions between parent and child could also help clinicians guide parents toward better learning strategies.
Gilkerson says researchers from dozens of universities have already used the tool to conduct their own studies on other issues too, such as the impact of television exposure on child development. “The LENA system can go far beyond this particular application of autism screening,” says Xu. “There’s tremendous opportunity for a lot of different applications in psychology research and language behavior research.”
The inside story of how ChatGPT was built from the people who made it
Exclusive conversations that take us behind the scenes of a cultural phenomenon.
How Rust went from a side project to the world’s most-loved programming language
For decades, coders wrote critical systems in C and C++. Now they turn to Rust.
Design thinking was supposed to fix the world. Where did it go wrong?
An approach that promised to democratize design may have done the opposite.
Sam Altman invested $180 million into a company trying to delay death
Can anti-aging breakthroughs add 10 healthy years to the human life span? The CEO of OpenAI is paying to find out.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.