US researchers have asked a group of families to film their children in interaction with objects and people. They tried eight models of automated learning to diagnose autism, which allows "simplifying the process and making it more effective," according to a study published in the scientific journal PLOS Medicine.
The study was developed by a team from the Faculty of Medicine at Stanford University and was led by Dennis Wool, a professor of pediatrics and biomedical data in that California city.
Each of the models contained a "set of algorithms that included 5 to 12 behavioral characteristics of children and gave a general result indicating whether the child has autism, "he explained.
How were the videos treated?
Wall said that to evaluate the models, they asked families who were recruited for the study to send extra videos of one to five minutes. which depicts the faces and hands of children and their "social interaction, as well as the use of toys, pencils and accessories" were captured.. Of these images, 116 boys with an average age of 4 years and 10 months were diagnosed with autism, and another 46 (with an average of two years and 11 months) developed, he explains.
Nine expert reviewers analyzed the videos with the help of 30 questions questionnaire with "yes" or "no" answers, based on the typical behavior of autism, which were then incorporated into the eight mathematical models.
The best-performing model is one that identifies 94.5% of cases of autistic children and 77.4% of those with children without autism. For verification of the results they evaluated 66 other videos, half of children with autism. The same model correctly identified 87.8% of cases of children with autism and 72.7% of those who did not have this disorder.
Another advantage of using home videos for diagnose is that they "take the child in their natural environment," as opposed to the clinical assessment carried out in a medium "which can be rigid and artificial and cause atypical behavior." "We have shown that we can identify a small group of behavioral characteristics that are highly consistent with clinical outcomes and that non-experts can evaluate these features quickly and independently in a virtual online environment within minutes," said Wol.