Home Uncategorized FDA authorizes machine learning software to help diagnose autism

FDA authorizes machine learning software to help diagnose autism


For the first time ever, the Food and Drug Administration gave the green light to an algorithm designed to help clinicians diagnose autism in young children. Developed by Palo Alto-based startup Cognoa, the software uses questionnaires from parents, physicians and home videos to make recommendations to assist pediatricians with diagnosis.

The goal is to identify autism spectrum disorder (ASD) first. On average, most children in the US are diagnosed around the age of 4.

“Many of these children are waiting long before reaching[to the specialist],” said Cogno CMO Dr. Sharif Tarman, a pediatrician, said in a Zoom interview. “It’s really a big deal. We haven’t found a way to get this kind of market authorization.”

Taraman said the software uses machine learning to identify “maximally predictive” features from questionnaires and two short home videos.

Of course, asking people to provide videos of their kids is very personal. He said families would need to allow video analysts and the physicians involved in their care to review the videos.

The FDA’s authorization was based on the results of a prospective, double-blind study that compared how much software helped diagnose autism compared to a panel of physicians diagnosed based on DSM-5 criteria. performed well. Cognoa went through the FDA’s new route for low- or medium-risk devices with no predicate.

It was assessed on 425 children ranging in age from 18 months to five years at 14 different sites. Taraman said the company also made it a point to recruit a diverse group of patients for testing in terms of race, ethnicity, gender, education and socioeconomic status. Currently, girls and minorities often diagnosed with ASD at a later age.

According to the FDA, Cognoa’s test resulted in a false-positive result in 15 out of 303 children tested without ASD. Meanwhile, it turned out to be a false negative in only one of the 122 children with ASD.

In cases where there was no clear diagnosis or rule-out, the algorithm gave an indeterminate result. Overall, it provided a diagnosis for about 32% of patients in the trial.

The ability to deliver an indeterminate result was important, Tarman said, so the algorithm wouldn’t give too many false positives, or overlook children who have other neurodevelopmental conditions that need to be addressed.

“Technology is always a tool. It should never be a replacement for a physician,” he said. “The trial is not meant to be a standalone.”

Cognoa plans to begin marketing the software, called Canvas DX, later this year.

“Autism is a really beautiful thing,” Taraman said. “Our goal is not to ‘shut down’ autism; we aim to address the challenges that come with autism.”

Photo Credit: DrAfter123, Getty Images



Please enter your comment!
Please enter your name here