Using AI to Revolutionize Child Behavioral Diagnostics and Therapeutics

At ODSC East’s opening day, experts from around the world gave presentations on how AI is breaking boundaries in their respective fields. One area that data scientists are hoping to reshape with the help of AI is the process of child behavioral disorder diagnostics.

Halim Abbas is the Chief AI Officer for Cognoa, a pediatric and behavioral health startup based in Palo Alto, Calif. looking to implement machine learning and artificial intelligence processes to categorize and diagnose disorders like Autism Spectrum Disorder and Attention Deficit/Hyperactivity Disorder in young children.

[Related article: Machine Learning and Compression Systems in Communications and Healthcare]

According to data from the Centers for Disease Control and Prevention, nearly one in 59 children are diagnosed with ASD, and 11 percent of children ages 4-17 have ADHD. The larger issue parents face prior to receiving a diagnosis for their children is the waiting game during which key developmental stages in a child’s life occur. Because there are no definitive medical tests or genetic markers for these disorders, medical professionals have to rely on clinical studies, which can be both expensive and time-consuming.

“The average age of Autism diagnoses in the United States remains at four years old, despite research that shows early intervention is beneficial,” Abbas said.

Abbas went on to explain that the different stages in a diagnosis can be lengthy, though they don’t have to be. In an attempt to create an easier pathway to treatment and intervention, Abbas proposes a holistic approach to diagnosing behavioral disorders. Abbas’ vision involves a questionnaire for parents to fill out that will be applied to a machine learning classifier, directed videos of the child which can be used for deep learning and visual analysis, interactive sessions involving games and storytelling, and finally an optional questionnaire for the child’s pediatrician which can be analyzed using machine learning algorithms. Ideally, this multi-module model would be faster, more accessible, and more reliable than current testing.

There are pitfalls to Cognoa’s model to consider, however, and Abbas is fully aware of them. Currently, U.S. citizens are in a stage where data privacy is an incredibly touchy subject—and rightfully so. In the past few years, healthcare providers have been the targets of security breaches, leaving patient data vulnerable. The Health Insurance Portability and Accountability Act prevents private companies from accessing large healthcare datasets without FDA approval, so Cognoa is currently working with limited data for its models. However, Abbas is hopeful for the future of AI in healthcare, which he sees as both secure and bright.

[Related article: Why Consumers Should Trust Companies with Their Data]

“I think the near future in the next 15 years will hold a role reversal between humans and machines,” Abbas said.

While current healthcare standards are resistant to change, Abbas sees the potential in big data analysis. Because disorders like Autism and ADHD exist on a spectrum, it’s difficult to differentiate between their varied degrees. Abbas explained that with millions of patient profiles and the different decision trees they might generate based on their diagnoses, we will be able to analyze highly dimensional data and create clusters, which could potentially help categorize disorders and learn about their differences and similarities in ways we simply cannot today.