American researchers have asked families to shoot their children who communicate with objects and people. They tried eight models of automated learning to the diagnosis of autism, which makes it possible to "streamline the process and make it more effective," studies published in the scientific journal PLOS Medicine.
The study was developed by a team at the Stanford University Medical School and led by Dennis Wall, professor of pediatrics and biomedical research at the University of California.
Each of the models contained a "set of algorithms that included 5 to 12 behavioral characteristics of children and generated a total score indicating whether a child has autism, "he explained.
How videos were processed
Wall said that family members who were admitted for the study wanted to send home videos for one to five minutes to evaluate the models. in which the faces and hands of the children were shown and their "social interaction, as well as the use of toys, pencils and dishes, were captured".From these images, 116 boys with an average age of 4 years and 10 months were diagnosed with autism, and another 46 (with a mean of 2 years and 11 months) developed, he explained.
Nine experts checked the video using a 30 questionnaire with "yes" or "no" responses based on the typical behavior of autism, which were then incorporated into eight mathematical models.
The model that offers the best results is the one that identified 94.5% of cases of autistic children and 77.4% of children without autism. To verify the results they have evaluated 66 more videos, half of them with autism. The same model correctly identified 87.8% of cases of autistic children and 72.7% of those who did not have this disorder.
Another advantage of using home videos for diagnostics is that they "take the child into their natural environment", unlike the clinical evaluation that is carried out in the medium "which can be stiff and artificial and causes atypical behavior". "We have shown that we can identify a small set of behavioral characteristics that are highly aligned with clinical outcomes, and that non-experts can quickly and independently evaluate these characteristics in a virtual online environment within minutes," Wall said.