Early detection of Autism is critical. Starting treatment at around 18 to 24 months can increase a child’s IQ up to 17 points. Early intervention can also equate to about $1.2 million saved in medical costs over the span of their life.
Currently, early screening relies on a questionnaire that parents fill out. However, it is inaccurate and often produces false positives. These results are hard to validate because there are few licensed ASD clinicians, so children often can’t receive a diagnosis until after they turn three. Researchers at Duke, however, aim to change this by creating an app using machine learning that could screen for ASD.
The app tracks the child’s gaze and facial expressions as they look at a screen with one social stimulus and one non-social stimulus. The researchers then use machine learning to turn the data into meaningful information. They trained the algorithm using input from a team of trained ASD clinicians until it was able to detect signs successfully.
Researchers tested their app and found that it was about 90% accurate, which is higher than the 50% accuracy of the currently used questionnaire. The app is also more efficient and could act as a complementary tool to licensed ASD clinicians.
This screening method is already being used to detect ADHD, but researchers say it could potentially be used to detect other behavioral disorders. Ultimately, this could make healthcare more accessible and would increase the efficiency and accuracy of early screening for Autism.
Sanjana is a junior in high school who is passionate about raising awareness about developmental disabilities and neurological disorders. This interest developed from neuroscience research she’s conducted and she wants to share information with the community.