When diagnosing autism spectrum disorder (ASD), parents and doctors cannot simply rely on a laboratory test that can identify a biochemical marker of the disorder. In fact, defining a person as autistic requires a lengthy process that includes observations of the person’s behavior, interviews with those responsible, and the use of tests to measure development. All of that could change in the coming years with the use of artificial intelligence (AI).
Facilitating the diagnosis of autism is the mission of an international team of researchers, including Brazilian, French and German scientists. The new identification method consists of the analysis of MRI scans performed by an algorithm. Published in the magazine Scientific Reports Preliminary data suggests that the strategy is promising.
How does the potential AI autism screening work?
“We started developing a method for collecting images using magnetic resonance imaging and electroencephalography,” explains Francisco Rodrigues, professor at the Institute of Mathematics and Computer Science at the University of São Paulo (ICMC-USP). FAPESP Agency.
A total of brain imaging data from 500 people were used, 242 of whom have an autism spectrum disorder. “By comparing maps from people with and without ASD, we found that it would be possible to diagnose using this methodology,” Rodrigues adds.
With the information obtained, the system was able to determine which changes in the brain are associated with autism. According to the researchers, the accuracy was over 95%.
Differences in the autistic brain
“Until a few years ago, little was known about the changes that lead to the onset of symptoms. For example, it is now known that altered brain circuitry in patients with ASD can be linked to certain behaviors,” says Rodrigues. In general, patients’ brain networks tend to be more disconnected, with less information distribution and less connectivity.
In fact, other studies have attempted to find out which brain changes are typical of the disorder. The problem is that the discrepancies are usually barely noticeable in current analyses, especially in milder cases.
Potential of AI in brain reading
In addition to autism, the brain reading and analysis method, with the necessary parameter adjustments, can also be used in the diagnosis of other diseases that affect the brain. For example, it should be used to detect cases of schizophrenia or Alzheimer’s.
However, experts emphasize that the tool for tracking cases of autism spectrum disorder is still in development and will require years of studies before it finally hits the market and doctors’ offices. However, the number of new diagnoses of childhood autism is increasing in the opposite direction, which should accelerate the process and attract more investment in the cause.