AI screens for fetal alcohol spectrum disorder with 85% accuracy

A new machine learning-based system, detailed in the journal Frontiers in Neurology, screens children for fetal alcohol spectrum disorder (FASD) in a quick and more cost-efficient way. The system was developed by researchers at the University of Southern California, Queen’s University in Ontario and Duke University, and will be accessible to children in more remote areas of the globe. 

FASD is a group of conditions related to drinking alcohol during pregnancy. Children with FASD can develop an abnormal appearance, short height, low body weight, poor coordination, behavior problems and impaired senses.

“Fetal alcohol spectrum disorders (FASD) is one of the most common causes of developmental disabilities and neurobehavioral deficits,” wrote first author Chen Zhang, a doctoral candidate from the Neuroscience Graduate Program at USC, and colleagues. “Despite the high-prevalence of FASD, the current diagnostic process is challenging and time- and money- consuming, with underreported profiles of the neurocognitive and neurobehavioral impairments because of limited clinical capacity.”

At present, the researchers noted, there is no blood test to diagnose FASD. The traditional method of diagnosis is subjective and involves various tests and clinical evaluations—which is costly. Co-researcher James N. Reynolds, PhD, of the Kids Brain Health Network, noted FASD is a “major public health problem” and the annual cost of it can range from $22,000 to $24,000 in the U.S. and Canada.

The researchers sought to develop a machine learning method that could screen for FASD in a cohort of 46 children exhibiting FASD. They determined a screening tool may be able to reach more children who are at risk for the conditions.

The tool uses a camera and computer vision to document FASD children’s eye movements while watching multiple one-minute videos and as they look toward or away from a specific object. The researchers compared those eye movements to a control group’s eye movements. Children with “outside the norm” eye movements were identified as potentially being at risk for FASD. In addition, the researchers utilized psychometric tests and neuroimaging of the corpus callosum, which bridges the left and right sides of the brain.

“The new screening procedure only involves a camera and a computer screen, and can be applied to very young children,” Zhang said in a prepared statement. “It takes only 10 to 20 minutes and the cost should be affordable in most cases. The machine learning pipeline behind this gives out objective and consistent estimations in minutes.”

Overall, the researchers found their method, which consists of three eye movement tasks and psychometric tests took an hour and a half to complete, with an average sensitivity of 82 percent, an average specificity of 88 percent and an accuracy rate of 85 percent.

This machine learning method can provide important feedback so parents can get help from professionals and FASD-affected children can receive early cognitive and behavioral interventions, the statement read.

“Such a screening procedure could be widely used at clinics, schools, or health units where young children are seen routinely, across different regions and areas, promoting communications within an interdisciplinary context,” the researchers noted.