Scientists create AI that can predict autism in babies

Scientists at the University of North Carolina have developed an algorithm that can diagnose autism in babies between 6 and 12 months old. Based on brain scans, the algorithm turned out to correctly predict the diagnosis in 81 percent of the cases.

This is according to a study published in the journal Nature. The accuracy of the 81 percent predictions is relatively high, according to IEEE Spectrum, as diagnoses of autism made based on behavioral research are usually only accurate in 50 percent of the cases. Moreover, these tests are usually not administered until the child is one year old.

The algorithm in the study used three factors: the surface area of ​​the child’s brain, the volume of the brain, and the child’s gender. The research shows that the algorithm was able to detect the increase in brain surface area of ​​children as young as six months old.

The researchers themselves are still skeptical about the practical usefulness of the algorithm. According to one of the researchers, it appears to be difficult and expensive to obtain brain scans from young children for research. In addition, the tests are not necessarily suitable for all children. Therefore, it seems unlikely that it is practically useful to be applied to every baby born.

Scientists know that the first signs that a child has autism can appear at a very young age, but it is still difficult to make a reliable diagnosis of autism in very young children. Previous studies have shown that people with autism have an increased brain volume and that this ‘overgrowth’ occurs early. The greater this ‘overgrowth’, the stronger the later symptoms of autism.