American researchers asked a group of families to film their children interacting with objects and people. They are trying eight models of automated learning to diagnose autism, which allows "simplification and process to make it much more efficient," according to the study published in the scientific journal PLOS Medicine.
The study was developed by a team of Stanford University School of Medicine and was led by Dennis Wall, a Pediatrician and Biomedical Data Science from that city in California.
Each of the models included "a set of algorithms including 5 to 12 children's behavioral characteristics and produced an overall score that indicated whether the child had autism, "he explained.
How the videos were handled
Wall said to evaluate the models, families recruited for the study were asked to send home videos from one to five minutes. where the faces and hands of the children were shown and their "social interaction as well as using toys, pencils and tools" were held.. Of these images, 116 boys with 4 years of age and 10 months of age were diagnosed with autism and another 46 (with an average of two years and 11 months) developed, he explained.
Nine expert reviewers analyzed the videos using a 30 question questionnaires with "yes" or "no" answers, based on typical behaviors of autism, which were then incorporated into the eight mathematical models.
The model that offered the best results is that which identified 94.5% of cases with children with autism and 77.4% of children with no autism. For a check of the results they have evaluated 66 other videos, half of children with autism. The same model correctly identified 87.8% of cases with children with autism and 72.7% of those who did not have this disorder.
Another advantage of using home videos for the diagnosed is that "take the child in their natural environment", unlike the clinical assessment that is conducted in a "that can be rigid and artificial and cause atypical behaviors". "We showed that we can identify a small group of behavioral characteristics that are very consistent with clinical outcomes and that non-experts can graduate these features quickly and independently in a virtual online environment, in minutes, "said Wall.