AI can spot signs of autism in babies as young as three months from brain scans
Brains scans could reveal signs of autism in children as young as three months, according to new research.
Using “low-cost, non-invasive” electroencephalograms (EEGs), scientists in Boston developed an algorithm that interprets EEG readings. The AI looks for patterns in the signals and can use these patterns to predict, or rule out, the chances of a child having autism spectrum disorder. EEGs use sensors placed on the scalp to measure brain activity and are typically used to diagnose epilepsy.
In a sample case of 188 children, the algorithm predicted autism spectrum disorder (ASD) in patients by the age of nine months with nearly 100% accuracy. It was also used to predict the severity of the conditon.
Due to the complex nature of ASD, diagnosing autism at any age can be difficult. Signs typically start to exhibit themselves at around the age of two, at which point autism can play a significant role in how children socialise and communicate.
READ NEXT: What is autism?
For the study, doctors and computational scientists from the Laboratories of Cognitive Neuroscience at Boston Children’s Hospital, Boston University and the University of San Francisco, studied data from the Infant Sibling Project, now called the Infant Screening Project.
This project was set up as a collaboration between Boston Children’s Hospital and Boston University with the aim of mapping early development, and identifying those infants at risk of developing ASD and/or language and communication difficulties. To work alongside this data, Dr William Bosl, associate professor of Health Informatics and Clinical Psychology at the University of San Francisco, has spent the past decade working on algorithms to interpret EEG signals. Bosl’s research suggests that even an EEG that appears “normal” contains “deep” data that unlocks the secrets of brain function, connectivity patterns and structure that can’t be seen by human practitioners.
The Infant Screening Project gave Dr Bosl EEG data from 99 children considered at high risk for ASD, because they have an older sibling with the diagnosis, as well as the readings from 89 low-risk children without an affected sibling. EEG readings were taken when the children were three, six, nine, 12, 18, 24 and 36 months old by fitting a net over the babies’ scalps with 128 sensors as the babies sat in their mothers’ laps. All babies also underwent behavioural evaluations with the Autism Diagnostic Observation Schedule (ADOS).
Bosl’s algorithms analysed six frequencies of the EEG – high gamma, gamma, beta, alpha, theta, delta – to measures the differences in how each of the babies’ brains were wired and how they processed information. The algorithms predicted a clinical diagnosis of ASD with “high specificity, sensitivity and positive predictive value” in over 95% of cases. Results which Bosl described as “stunning.”
The predictive accuracy between three months and nine months of age was almost 100% and the team was able to predict ASD severity, as indicated by the ADOS Calibrated Severity Score, with “quite high reliability” also by the age of nine months.
“EEGs are low-cost, non-invasive and relatively easy to incorporate into well-baby checkups,” said Dr Charles Nelson, director of the Laboratories of Cognitive Neuroscience at Boston Children’s Hospital and co-author of the study. “Their reliability in predicting whether a child will develop autism raises the possibility of intervening very early, well before clear behaviuoral symptoms emerge. This could lead to better outcomes and perhaps even prevent some of the behaviors associated with ASD.”
Bosl believes that the early differences in signal complexity, on multiple aspects of brain activity, fit with the view autism is a disorder that begins during the brain’s early development but can take different routes.
The research is published in the journal Scientific Reports.