By scanning the growing brains of 148 babies at high and low risk for autism, researchers were able to predict which children would develop the disorder within the first year of life before symptoms began to appear, and the diagnosis was made at two o’clock, researchers reported Wednesday in Nature.
Those researchers, led by psychologists Heather Cody Hazlett and Joseph Piven of the University of North Carolina at Chapel Hill, came up with the idea for the study after previously finding that children with autism tend to have larger brains than children without the disorder. To follow up on this, they used magnetic resonance imaging to track and predict brain overgrowth as it happened. All in all, the study raises hopes that doctors will one day be able to make rapid diagnoses, enabling them to intervene ever earlier.
The study has limitations, of course: It was small, so researchers will have to repeat it with many more children to confirm the findings. It also only applies to babies who are at high risk of developing autism, which are babies who already have siblings diagnosed with the disorder. For families with one child with autism, there is about a one in five chance that subsequent children will also be affected. In the general population in the US, autism is diagnosed in about one in 68.
Still, “the findings laid the groundwork for the field to try to implement interventions before the symptoms that define autism consolidate into a diagnosis,” Jed Elison, a child development expert at the University of Minnesota and a co-author of the study, said in a statement. pronunciation.
For the study, researchers took MRI scans of 106 high-risk babies at 6, 12 and 24 months. They did the same for 42 low-risk babies. As before, they noted that children later diagnosed with autism at 24 months had larger brains. More specifically, their brain’s cortical surface — the folded, corrugated outer layer of the brain — grew faster in the first 6 to 12 months compared to people without a diagnosis.
The researchers then used the early scans and machine learning to develop an algorithm that could predict the development of autism. In one test, the algorithm correctly predicted 30 of 37 autism diagnoses, an accuracy rate of 81 percent. For a group of 142 who had not been diagnosed with autism, the algorithm incorrectly predicted that only four of them would be diagnosed.
The researchers are optimistic but cautious. False positive diagnoses can be devastating to a family, and the technology is still in its infancy. But with more brain scans and data from other forms of imaging, the researchers hope that the predictions will become stronger and more widely applicable.
Nature2017. DOI: 10.1038/nature21369 (About DOIs).