AI Tool to Diagnose Autism Could Give Concerned Parents a Fast Diagnosis

The startup Cognoa has submitted its app-based tool for FDA clearance

4 min read
Caregivers record child’s natural behavior at home through Cognoa’s parent-facing mobile app, one component of Cognoa’s autism diagnostic device.
Caregivers record child’s natural behavior at home through Cognoa’s parent-facing mobile app, one component of Cognoa’s autism diagnostic device.
Photo: Cognoa

This week, a California-based company announced it will seek FDA clearance for a first-of-its-kind autism spectrum disorder (ASD) diagnostic tool. Cognoa’s technology uses artificial intelligence to make an ASD diagnosis within weeks of signs of concern—far faster than the current standard of care. If cleared by the FDA, it would be the first tool enabling primary care pediatricians to diagnose autism.

The approach is “innovative,” says Robin Goin-Kochel, a clinical autism researcher at Baylor College of Medicine and associate director for research at Texas Children’s Hospital’s Autism Center, who is not affiliated with Cognoa. The field absolutely needs a way to “minimize the time between first concerns about development or behavior and eventual ASD diagnosis,” she adds.

Cognoa’s tool is the latest application of AI to healthcare, a fast-moving field we’ve been tracking at IEEE Spectrum. In many situations, AI tools seek to replace doctors in the prediction or diagnosis of a condition. In this case, however, the application of AI could enable more doctors to make a diagnosis of autism, thereby opening a critical bottleneck in children’s healthcare.

While most parents of children with autism notice developmental changes early on, within the first 1 to 3 years of life, the median diagnosis age in the United States is 4.3 years old. That’s because families often wait months, even years, to see a specialist and get a diagnosis. The time lost during that period is critical: Numerous studies show that early intervention, before the syndrome is fully manifest, can reduce the severity of ASD and improve a child’s brain and behavioral development.

Cognoa’s technology comes out of the lab of founder Dennis Wall, an associate professor of pediatrics at the Stanford University School of Medicine. “I went into this with the hope of objectively asking the question: Can we reduce the complexity of the autism diagnostic process without loss of accuracy?” says Wall.

By feeding electronic health record data into a set of algorithms, Wall’s team was able to distinguish particular features central to an ASD diagnosis, including social and emotional traits such as smiling in response to another person’s smile, joint attention at an object, creativity, and imagination.

The team’s ASD diagnostic seeks to capture those features with three modules: a parent survey, home videos, and a clinician questionnaire.

David Happel, CEO of Cognoa, explains how the tool works: When a parent expresses concern at a pediatrician appointment, or a child fails an ASD screening questionnaire, the pediatrician gives the parent a code to access Cognoa’s app on their smartphone. Once in the app, the parent answers a 15-minute questionnaire about their child’s behavioral patterns, then uploads two home videos of the child, 1 to 2 minutes in length, capturing the child’s behavior in a natural environment. The videos are sent to a trained Cognoa professional who reviews them and answers pertinent questions. Those answers are fed into Cognoa’s AI along with the parent answers and a short questionnaire filled out by the pediatrician. Then, the algorithm sends a result to the pediatrician, and the pediatrician renders a diagnosis.

The tool’s algorithms are trained on data from hundreds of real cases across genders, races and ethnic backgrounds, says Happel. “It has proven to not only accelerate the time of diagnosis, but also remove a lot of the biases that are inherently in place in the current system.” Today’s standard ASD diagnostic tools, Happel notes, were constructed with health data from young Caucasian boys, so girls and children of non-white backgrounds are not well recognized by the tools, contributing to delays in diagnosis for those groups.

Today’s standard ASD diagnostic tools were constructed with health data from young Caucasian boys, so girls and children of non-white backgrounds are not well recognized by those tools.

In a Cognoa study published in March, which demonstrated an earlier version of the tool, the ASD diagnostic out-performed current screening tools for autism. Today's screening tools include questionnaires answered by a parent, teacher, or clinician that flag some at-risk children.

More recently, the company completed a pivotal, double-blind clinical trial at 14 sites around the United States. The trial involved 425 participants, aged between 18 and 72 months, whose parents or doctors had expressed concern about their development but had not previously been evaluated for ASD, according to a press release. Each child was assessed twice: once using Cognoa’s tool, and once by a specialist clinician based on DSM-5 criteria, whose diagnosis was validated by a second specialist clinician.

The results of the pivotal trial are not yet published, so there is no specific data to report, but the company says the trial “surpassed its targeted benchmarks”—as agreed upon with the FDA—and was accurate across genders and races. Uniquely, the study ran from July 2019 through May 2020, so some of the children were evaluated remotely this spring during the pandemic via telemedicine. The tool performed equally as well when administered remotely, says Happel.

The company plans to submit the full study for publication in coming months. And formal submission to the FDA will occur shortly, says Happel. He hopes to receive approval in the second half of 2021, then be ready to launch the product into the hands of pediatricians two months after that.

If and when Cognoa’s technology or others become available, “it will be really important to understand the plan for how new tools and technologies will be implemented in primary pediatric care,” says Goin-Kochel. Doctors are often slow to adopt new models, especially for making diagnoses they may not feel comfortable making, she notes, and new technologies raise practical questions such as when they should be applied and whether insurance companies will pay.

“I’m very hopeful there is a near-term future where this product is available, covered by insurance, and made available to everybody as immediately as possible,” says Wall. He is now working on several technology-assisted therapies for ASD, including a project with Cognoa using Google Glass as part of a behavioral therapy to help children with autism recognize emotion.

This article appears in the November 2020 print issue as “AI Diagnoses Autism Early.”

The Conversation (0)

This CAD Program Can Design New Organisms

Genetic engineers have a powerful new tool to write and edit DNA code

11 min read
A photo showing machinery in a lab

Foundries such as the Edinburgh Genome Foundry assemble fragments of synthetic DNA and send them to labs for testing in cells.

Edinburgh Genome Foundry, University of Edinburgh

In the next decade, medical science may finally advance cures for some of the most complex diseases that plague humanity. Many diseases are caused by mutations in the human genome, which can either be inherited from our parents (such as in cystic fibrosis), or acquired during life, such as most types of cancer. For some of these conditions, medical researchers have identified the exact mutations that lead to disease; but in many more, they're still seeking answers. And without understanding the cause of a problem, it's pretty tough to find a cure.

We believe that a key enabling technology in this quest is a computer-aided design (CAD) program for genome editing, which our organization is launching this week at the Genome Project-write (GP-write) conference.

With this CAD program, medical researchers will be able to quickly design hundreds of different genomes with any combination of mutations and send the genetic code to a company that manufactures strings of DNA. Those fragments of synthesized DNA can then be sent to a foundry for assembly, and finally to a lab where the designed genomes can be tested in cells. Based on how the cells grow, researchers can use the CAD program to iterate with a new batch of redesigned genomes, sharing data for collaborative efforts. Enabling fast redesign of thousands of variants can only be achieved through automation; at that scale, researchers just might identify the combinations of mutations that are causing genetic diseases. This is the first critical R&D step toward finding cures.

Keep Reading ↓ Show less