new early diagnosis: EEGs

February 23, 2011 — A test that combines electroencephalograms (EEGs) and machine-learning algorithms may help identify developing features of an autism spectrum disorder (ASD) in children younger than 1 year, new research suggests.


In the pilot study, investigators found that the EEG-based test had 77% accuracy in distinguishing between 9-month-old infants known to be at high risk for autism, due to having an older sibling diagnosed as having an ASD, and their healthy counterparts.


Although the between-groups accuracy rate was even higher when looking specifically at boys at 9 months of age and stayed significant at their 18-month mark, the highest rate of class accuracy for girls was when they were 6 months of age.



Dr. William Bosl

"Using algorithms to find distinctive patterns in the EEG signals, we were able to pick out which kids were in the control group and which ones were in the high-risk group," lead study author William Bosl, PhD, research scientist at Children's Hospital Boston and instructor in pediatrics at Harvard Medical School, Boston, Massachusetts, told Medscape Medical News.


"We believe this is telling us there's an endophenotype or characteristics of autism that are carried on in the family and are particularly evident at 9 months in the brain's electrical activity," explained Dr. Bosl. Autism, he added, is not usually diagnosed before the age of 3 years.


"We know that early intervention leads to better outcomes with reduced symptoms and being able to better interact with society. Right now, 'early' means at 3 years. If we could start therapy at 9 months or 12 months, before the symptoms even occur, it is believed that the outcomes could be even better — although we don't know that just yet," said Dr. Bosl.


He noted that although validation is needed for the study findings, this is a quick, safe, inexpensive, and practical way of capturing very early differences in brain organization and function.


"If this research leads to an early diagnostic method based on EEG alone, that could have tremendous implications for kids who have autism."


The study was published online February 22 in BMC Medicine.


Recent Technical Advancements


"A great deal of information about interrelationships in the nervous system likely remains undiscovered because the linear analysis techniques currently in use fail even to detect them," write the researchers.


Dr. Bosl added that although EEGs have been around a long time and have been primarily used by neurologists to study epilepsy and seizures, recent developments "in the physics of complex systems" have led to being able to now see new features in the EEG's signals that might give more information about how the brain is functioning.


"Also, recent improvements in computer science now enable us to find complicated patterns in data that human eyes can't see. That, plus advances in neuroscience that tell us autism is a disorder that develops over time, led to this study.


"We wanted to track a group of children by looking at the changing features in their EEG signals over time to see if we could find distinctive patterns that would allow us to discriminate autism characteristics developing long before they could be observed. In a nutshell, we were looking for a practical way of diagnosing autism," said Dr. Bosl.


The investigators evaluated data on a cohort of 79 infants who were participating in the Infant Sibling Project study. Of these, 46 had an older sibling with an ASD diagnosis (infant sibs), whereas 33 had no family history of ASDs (healthy controls).


"Typically, about 20% of the infant sibs will go on to have a diagnosis of autism, around 40% to 50% of them will not have that diagnosis but will have some characteristics that are shared with autism, and the remaining will not develop autism at all," said Dr. Bosl.


Resting EEG signals were recorded for all participants with a 64-point Sensor Net System starting at 6 months and repeated when possible when they were 9, 12, 18, and/or 24 months of age.


The investigators then computed the modified multiscale entropy (mMSE) of each EEG brainwave channel to explain the density of neurons in each part of the brain, how the connections between them were organized, and the balance of short- and long-distance connections.


New Psychiatric Biomarker?


Results showed that mMSE "appears to go through a different developmental trajectory in infants at high risk for autism than it does in typically developing controls, [and] differences appear to be greatest at ages 9 to 12 months," report the researchers.


They note that at 9 months, infants undergo important developmental milestones, including the emergence of higher-level social and communication skills — which are often impaired in those with ASDs.


For boys, the test's classification accuracy was almost 100% at the age of 9 months. Although it decreased to 90% at 18 months, the accuracy rate was still significant. The 70% accuracy rate at 12 months was not deemed significant.


For girls only, the accuracy rate was 80% at 6 months of age. However, the rate decreased and was no long significant at any of the later time points.


The study authors note that this could possibly indicate "a sex difference in developmental trajectories."


They add that more studies are now needed and that a "deeper understanding of the relationship between neurophysiological processes and cognitive function may yield a new window to the mind and provide a clinically useful psychiatric biomarker."


Dr. Bosl reports that the investigators plan to continue following up the infant sibs group over time and to compare EEG patterns in those who receive an actual ASD diagnosis to those who appear to be developing normally — and then compare all outcomes to the healthy controls.


In addition, the researchers are collecting data from older children between the ages of 6 and 17 years to eventually compare their EEG patterns for different types of ASDs.


"Overall, I'd tell clinicians to keep watching. I'm building a team to push this work forward as quickly as possible, not just as a research project but as a practical clinical tool," said Dr. Bosl.


"With some refinement, here's a device that's cheap enough and easy enough to be used by pediatricians. There really is the possibility here of developing a simple diagnostic tool for measuring kids' cognitive development and detecting things like autism."


The study was funded by grants from Autism Speaks, the National Institute on Deafness and Other Communication Disorders, and the Simons Foundation. Dr. Bosl reports being named on a provisional patent application that includes parts of the signal analysis methods used in the study. The other study authors have disclosed no relevant financial relationships.


BMC Med. Published online February 22, 2011.