Can An Algorithm Predict Autism?
20 Feb 2017
A new study from Carolina Institute of Developmental Disabilities at the University of North Carolina suggests that by using MRI scans, a computer program and an algorithm, autism can be predicted in high-risk infants. If these findings hold up, it would signal that new diagnostic innovations could allow earlier intervention for children who will more likely be diagsoned with ASD (Autism Spectrum Disorder).
Here, Medical News Today (MNT) described the study that was published in Nature on February 16, 2017:
In their search for biological markers of ASD in developing children, the team analyzed MRI scans at 6 and 12 months of age. They compared the scans of three groups:
- Infants with an older sibling with autism (high-risk) who went on to develop autism after the age of 2
- Infants with an older sibling with autism who did not develop autism after the age of 2
- Infants with a low family risk who did not go on to develop autism.
The researchers measured surface area, brain volume, and cortical thickness. By plugging this data into a computer program, the team designed an algorithm to predict whether an infant would later develop ASD. They then tested the accuracy of this algorithm in a separate trial.
The brain differences as computed by the algorithm were shown to correctly predict 8 out of the 10 children who would later go on to develop autism.
Although the study targeted these high-risk children, more extensive studies, techniques and diagnostic tests can expand this studies accuracy to all infants.
This potential breakthrough still leaves us without the causes of ASD, but it provides even more evidence that it is caused–at least, in part–by genetics factors. Additionally, this study gives us hope for earlier detection, which will lead to better treatments and therapies of symptoms.