Pushing diagnostic capacity out of childhood into infancy: neuroimaging study spots signs of autism in high-risk babies
Février 18, 2017
Until very recently, no one has been able to accurately predict autism in babies under a year of age.
That picture has begun to change, as discoveries arising from genetics, nuanced parental observation and neuroimaging are all pointing in the same direction: the capacity to identify autism spectrum disorder in children before overt symptoms – such as loss of, or delays in speech and lack of eye contact – begin to show around the age of two.
The overall sense is one of excitement,” says Dr. Lonnie Zwaigenbaum, co-author of “Early brain development in infants at high risk for autism spectrum disorder,” a neuroimaging study published in Nature that is receiving international attention.
“The paper definitely tells a new story that expansion of brain surface area prior to the first birthday precedes overall increases in brain growth in the second year - and that the particular pattern is highly predictive of risk of ASD in younger siblings - even without taking account of behavioural symptoms, adds Dr. Zwaigenbaum.
The findings were more consistent with a developmental mechanism than injury/inflammation, which may imply that genes governing brain growth and development were involved.
Zwaigenbaum was part of a US-Canada research team that scanned the brains of 106 babies whose siblings have autism at the ages of 6, 12 and 24 months using MRI. Forty-two low-risk infants were also followed in the successful attempt to document differences in brain development during this critical period.
Since the 1990s, researchers have noted that children with ASD tend to have larger head sizes, likely caused by brain overgrowth. Zwaigenbaum and colleagues captured differences in both brain volume and cortical surface area – a measure of folds that cover the brain – between the 15 children in the study who went on to be diagnosed with autism at age two, and those who did not receive a diagnosis.
Using a combination of MRI brain scans and a deep-learning neural network – a variation of machine learning – to analyze their data, the researchers found their algorithm identified 37 infants as being likely to have autism. Thirty of these children were ultimately diagnosed – meaning the analysis was accurate for 81% of the study participants. Only four of the 142 infants that lacked this brain growth pattern were diagnosed with autism.
“Signs of autism in children prior to age 12 months are subtle and inconsistent,” observes Dr. Zwaigenbaum, who is a clinician as well as a researcher. “It can be hard to find differences at six months, even though they may be pronounced at a year.”
Those early subtle signs – reactivity, motor differences, differences in attention – may speak most loudly to parents who already have a child with autism, Zwaigenbaum notes. Recent papers, including a study assessing the accuracy of a screening tool for parents reinforce that asking parents about their concerns also remains a powerful strategy for early detection.
“It’s an open question how behavioural observations and biomarker data from imaging and genetics correlate,” says Zwaigenbaum. “Do they identify different kids? Do biomarkers improve your predictions?”
While the ability to push diagnosis ever earlier into infancy is exciting – with potentially game-changing effects on developmental trajectories and life course – those outcomes depend on the availability of, and access to effective interventions. “There are issues we have to grapple with ethically, in pushing the boundaries of early detection,” Zwaigenbaum reflects. “If we push the conversation to as early as six months, what could we offer to parents?”
At this point, adds Zwaigenbaum, “all we can confidently say is our findings are predictive amongst younger siblings. As well, brain scanning would be a very expensive approach, even for a targeted group.”
The study has broader implications. “It encourages us to look for less expensive strategies, including eye tracking. We also need to ask parents about their concerns, and to integrate findings from different areas of research.
“We need intervention strategies that are suitable,” concludes Zwaigenbaum. “But we are making progress. Until recently, we weren’t even able to think about intervention prior to the age of two, and now we have promising evidence-based approaches that can be started shortly after the first birthday. We just have to continue to keep pace.”
NeuroDevNet est financé par les Réseaux de centres d’excellence, programme du gouvernement fédéral pour avancement de la science et de la technologie.