American researchers have asked a family group to shoot their children by interacting with objects and people. They tried eight automated learning models to diagnose autism, which allows "streamline the process and make it more effective," according to a study published in the scientific journal PLOS Medicine.
The study was developed by the Stanford University School of Medicine, headed by Denss Walls, Professor of Pediatrics and Biomedical Data Science, California.
Each of the models had a "series of algorithms that Includes 5 to 12 behavioral characteristics for children and it produced a general result indicating whether the child had autism, "he explained.
How videos were processed
Siena said that in order to evaluate the models, they asked the families who were hired to send home videos for one to five minutes. which featured children's faces and hands, as well as their "social interactions, as well as the use of toys, pencils and dishes".. Of these images, 116 children 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 it, he explained.
Nine expert reviewers analyzed video clips using 30 questionnaires with the answer "yes" or "no" based on the typical principles of autistic behavior, which were then included in eight mathematical models.
The best result was a pattern that identified 94.5% of cases in children with autism and 77.4% of those without autism. To test the results they rated 66 other videos, half of children with autism. The same model correctly identified 87.8% of children with autism and 72.7% of those who did not have this disorder.
Another advantage of using home videos diagnosis is that "they reach the child in a natural environment", unlike the clinical evaluation that is done in the environment, "which can be rigorous and artificial and cause atypical behavior". "We showed that we can identify a small group of behavioral patterns that is highly consistent with clinical outcomes and that experts who are not experts can quickly and independently evaluate these properties in a virtual online environment in a matter of minutes," Wall said.