Artificial intelligence has now made its way toward accurately predicting the chances of a baby developing autism spectrum disorder (ASD) later in life.
In a first, researchers used brain scans of 6-month-old babies to spot changes in how specific regions of infant brains connect and synchronize, which can eventually predict if they’re at an elevated risk of developing autism at age 2.
Autism Spectrum Disorder
The new study led by teams from the University of North Carolina as well as Washington University follows an earlier project that used MRIs taken at both 6 and 12 months to arrive at similar predictions. Its new technique slashed the amount of scans necessary to produce the forecast, as well as predicted with over 96 percent accuracy versus the 81 percent of the earlier research.
ASD is an umbrella term for a number of impaired social and communication functions, which usually surface at about age 2 and with most kids diagnosed at age 4.
"The more we understand about the brain before symptoms appear, the better prepared we will be to help children and their families,” said senior author Dr. Joseph Piven in a statement.
Co-senior author Dr. John R. Pruett Jr. added that while there are no behavioral features to help identify autism before the symptoms appear, brain scans could identify ultra-high-risk children for earlier intervention.
Study Details
The team probed the brain scans of 59 so-called baby sibs, which are kids with an older sibling with autism. Baby sibs have a one in five chance of developing autism compared to the one in 68 in the general population.
In the study, the babies were placed in an MRI machine and scanned to analyze the neural activity of 230 brain regions. Instead of investigating anatomical differences, the scientists measured how the different areas connected and synchronized with each other, reflecting important coordination for functions like memory, cognition, and behavior.
Of more than 26,000 potential pairs of connections, they saw 974 in the 6-month-old babies that could be useful in predicting a diagnosis by age 2. They fed the information into a machine learning program, and saw astounding results.
"When the classifier determined a child had autism, it was always right,” shared author Robert Emerson. “But it missed two children. They developed autism but the computer program did not predict it correctly, according to the data we obtained at six months of age."
The authors emphasized that their small study has to be replicated, as a larger sample will further gauge the accuracy of the technique.
The brain technology used in the research is relatively expensive at several thousand dollars. Yet more affordable and non-invasive methods, including EEG or near-infrared spectroscopy, could one day be used in the clinical environment.
Future Of Autism Diagnosis
Psychiatry and neuroscience professor Joy Hirsch of Yale University, who wasn’t involved in the study, dubbed it a “really, really important advance” even though this version might not be the clinical standard in the future.
“[I]t paves the way for earlier and objective diagnosis,” Hirsch said.
Emerson said the most groundbreaking work is still forthcoming, and that using data in infant brains together could be the future of biological diagnostics for the disease during infancy.
The findings were detailed in the journal Science Translational Medicine.
