Screening for Autism Is AI Proving Ground in Pediatrics

April 24, 2025

Artificial intelligence is accelerating the early detection of developmental conditions like autism spectrum disorder (ASD), potentially allowing children to begin therapy earlier and significantly improving their long-term outcomes.

That has made ASD a proving ground for the use of AI in pediatric medicine. Numerous pilot projects have shown promising results and hope that machine learning techniques can be applied to other conditions.

Why It Matters

The parents of children with ASD tend to notice developmental challenges some time before the child turns 3 years old. However, there are frequently lengthy delays in testing and diagnosis. In many countries, the typical age of diagnosis is between 5 or 6 years old.

On top of that, ASD tends to be diagnosed through questionnaires rather than blood testing or genetic testing. As a result, there are concerns about standardization of the responses and children are frequently misdiagnosed.

“AI excels at identifying subtle behavioral and genetic patterns that humans may overlook due to cognitive biases or limitations, ensuring more consistent and objective early diagnoses,” said IEEE Senior Member Dheerendra Panwar.

Emerging Methods

Researchers have used a variety of methods to detect ASD. One method used sensors and imaging to detect changes in speech and language skills. IEEE Fellow Shrikanth Narayanan was awarded the IEEE 2025 James L. Flanagan Speech and Audio Processing Award for that line of research.

Brain scans have also emerged as a screening tool. One group of researchers used machine learning to determine that children with autism applied different amounts of force and moved their fingers differently while playing touchscreen games, providing yet another avenue for diagnosis. Because that research is noninvasive, other researchers furthered touchscreens as a diagnostic tool and invented ways to distinguish autism from other, closely related conditions.

Next Steps

While the research has been promising, most of the results are still in the experimental phase and little has been done in clinical settings.

So far, one smartphone app has gained FDA approval as a diagnostic screening device for autism, though it should be used in conjunction with other medical information.

But the success of AI-driven autism studies have driven hope for diagnostic tools in other areas, like dyslexia, attention deficit hyperactivity disorder and depression, according to a paper found in IEEE Xplore.

Research in this area continues. If upcoming larger trials continue to show accuracy and efficiency, AI-driven developmental assessments could become a normal part of 18- or 24-month well-child visits.