Developing an artificial intelligence algorithm that facilitates the process of detecting autism by 95 percent

Written By Mark

The Russian Science Foundation announced that scientists from Russia and China have managed to develop an artificial intelligence algorithm that will facilitate the process of detecting autism by 95 percent, by analyzing electroencephalogram data.
“The algorithm we developed specifically aims to search for the distinctive features of children with autism, and compare their EEG data with data from healthy children,” said Professor Alexander Khramov of the Baltic State University in Russia. “This algorithm will enable us to use machine learning for artificial intelligence to quickly detect symptoms of autism in its early stages.”
Khramov noted that scientists developed the algorithm using a special type of artificial intelligence technology called a variational autoencoder. The algorithm was programmed to detect specific patterns in data obtained from electroencephalograms. To do this, they collected electroencephalogram data from 298 children aged 2 to 16, half of whom had various forms of autism. The data from children with the disease was compared with data from healthy children.
Tests conducted on the algorithm after its development showed that it was able to identify people with autism based on EEG data with 95 percent accuracy, and it did not give any false positives, meaning that it did not classify healthy people as having autism.
In addition, the algorithm allowed scientists to identify several hallmarks of autism, including weak functional connections in the frontal lobe of the brain, so its developers suggested that it and EEG data could be used to develop new mechanisms for detecting the disease.