Calibrated Severity Index (CSS), part of the Diagnostic Autism Diagnostic Chart (ADOS), measures the severity of ASD symptoms based on clinical evaluation. This figure illustrates the CSS estimates that were predicted for each participant, based on machine learning algorithms applied to EEG data.


Credit: William Bozl, PhD.

Autism is more difficult to diagnose, especially at an early age. A new study in the Scientific Reports magazine shows that inexpensive EEGs that measure electrical brain activity accurately predict or exclude autism spectrum disorder (ASD) in infants, in some cases even at the age of 3 months.

"EEGs are inexpensive, non-invasive, and relatively easy to implement in child surveys," says Charles Nelson, MD, director of the Cognitive Neuroscience Laboratory at Boston Children's Hospital and co-author of the study. "Their reliability in predicting whether child autism develops increases the likelihood of early intervention, long before the onset of obvious behavioral symptoms, which can lead to better results in the treatment of autism and possibly even prevent some of the symptoms associated with ASD." >

The study analyzed data from the Children's Nursing Project (now called the Children's Screening Project), cooperation between the Boston Children's Hospital and the University of Boston, which aims to develop the map early and identify babies at risk for developing ASD and / or linguistic and communication difficulties.

William Bozle, Ph.D., associate professor of medical informatics and clinical psychology at the University of San Francisco, also affiliated with the Informatics in Computing Medicine Program (CHIP) at the Boston Children's Hospital, has been working on algorithms for the interpretation of EEG signals for almost a decade . Bozle's studies show that even a normal EEG contains "deep" data that reflects the brain function, connectivity patterns and structure that can only be found with computer algorithms.

The children's screening project provided EEG data from 99 infants considered at high risk for ASD (having an older brother with a diagnosis) and 89 with a low risk level (without an older brother). EEG were taken at the age of 3, 6, 9, 12, 18, 24 and 36 months, setting a grid over the heads of infants with 128 sensors, when the children were sitting on their mother's lap. All children also underwent extensive behavioral assessments using the Diagnostic Diagnostic Autism Diagnostic (ADOS), an established clinical diagnostic tool.

Six different components (frequencies) of the EEG (high gamma, gamma, beta, alpha, theta, delta) were analyzed in Bozl's computational algorithms using a variety of signal complexity measures. These measures can reflect differences in how the brain is connected and how it processes and integrates information, says Bozl.

The algorithms predicted a clinical diagnosis of an autism spectrum disorder with high specificity, sensitivity and positive prognostic value exceeding 95 percent in some age groups.

"The results were stunning," says Bozl. "Our projected accuracy for 9 months was almost 100 percent. We also were able to predict the severity of the ASD, as evidenced by the ADOS calibration rate with rather high reliability. "

Bozle believes that early differences in signal complexity based on multiple aspects of brain activity are consistent with the fact that autism is a disorder that begins during early brain development but can take different trajectories. In other words, an early predisposition to autism symptoms may depend on other factors along the way.

"We believe that babies who have an older brother with signs of autism can be genetically responsible for developing autism," says Nelson. "This increased risk may interact with another genetic or environmental factor, leading to some children developing autism, although clearly not all, because we know that four out of five babies do not develop autism."

History Source:

Materials provided by the Boston Children's Hospital. Note. Content can be edited for style and length.

Journal Reference:

William J. Bosle, Helen Tager-Flussberg, Charles A. Nelson. EEG-analytics for the early detection of an autistic spectrum: a data-based approach. Scientific reports, 2018; 8 (1) DOI: 10.1038 / s41598-018-24318-x