Autism Predictor

Machine learning and artificial intelligence (AI) have become valuable tools for diagnosing illness because of their ability to analyze and correlate huge amounts of data. In a recently published study, researchers have applied this concept to predict whether high-risk infants (i.e., younger siblings of children with autism spectrum disorder or ASD) are likely to develop autism.

In the study, functional magnetic resonance imaging (fMRI) was used to capture the brain activity of 59 six-month-olds, all of whom had an older sibling already diagnosed with ASD. A total of 230 brain regions were scanned, resulting in more than 26,000 neural connections that were analyzed for unique patterns. The data was used to create a machine-learning algorithm that searched for those patterns and compared the results to behavioral tests conducted when the same subjects were two years of age. The algorithm correctly identified nine of the eleven children who ultimately developed autism.

Methods such as this may someday revolutionize healthcare by enabling doctors to develop early preventative interventions for ASD and other diseases or disorders. Identifying risk factors early and with greater accuracy generally results in more positive outcomes. In addition, less invasive diagnostic procedures can improve accessibility of care in areas of the world where availability of hospitals and healthcare professionals is limited.