A new video analysis tool may help spot early signs of autism. The software, which is currently under development as a tablet application, tracks and records infants’ activity during videotaped autism screening tests.
Results show the program is as effective at spotting behavioral markers of autism as experts giving the test themselves, and the software performs better than non-expert medical clinicians and students in training.
“We’re not trying to replace the experts,” says Jordan Hashemi, a graduate student in computer and electrical engineering at Duke University.
“We’re trying to transfer the knowledge of the relatively few autism experts available into classrooms and homes across the country. We want to give people tools they don’t currently have, because research has shown that early intervention can greatly impact the severity of the symptoms common in autism spectrum disorders.”
The study focused on three behavioral tests that can help identify autism in very young children.
In one test, an infant’s attention is drawn to a toy being shaken on the left side and then redirected to a toy being shaken on the right side. Clinicians count how long it takes for the child’s attention to shift in response to the changing stimulus. The second test passes a toy across the infant’s field of view and looks for any delay in the child tracking its motion. In the last test, a clinician rolls a ball to a child and looks for eye contact afterward—a sign of the child’s engagement with their play partner.
In all of the tests, the person administering them isn’t just controlling the stimulus, he or she is also counting how long it takes for the child to react—an imprecise science at best. The new program allows testers to forget about taking measurements while also providing more accuracy, recording reaction times down to tenths of a second.
“The great benefit of the video and software is for general practitioners who do not have the trained eye to look for subtle early warning signs of autism,” says Amy Esler, an assistant professor of pediatrics and autism researcher at the University of Minnesota, who participated in some of the trials highlighted in the paper, which will appear online in the journal Autism Research and Treatment.
“The software has the potential to automatically analyze a child’s eye gaze, walking patterns, or motor behaviors for signs that are distinct from typical development,” Esler says. “These signs would signal to doctors that they need to refer a family to a specialist for a more detailed evaluation.”
According to Hashemi and his adviser, Guillermo Sapiro, professor of electrical and computer engineering and biomedical engineering at Duke, because the program is noninvasive, it could be useful immediately in homes and clinics. Neither, however, expects it to become widely used—not because clinicians, teachers, and parents aren’t willing, but because the researchers are working on an even more practical solution.
Later this year, the Duke team plans to test a new tablet application that could do away with the need for a person to administer any tests at all.
The program would watch for physical and facial responses to visual cues played on the screen, analyze the data and automatically report any potential red flags. Any parent, teacher, or clinician would simply need to download the app and sit their child down in front of it for a few minutes.
The efforts are part of the Information Initiative at Duke, which connects researchers from disparate fields to experts in computer programming to help analyze large data sets.
“We’re currently working with autism experts at Duke Medicine to determine what sorts of easy tests could be used on just a computer or tablet screen to spot any potential concerns,” says Sapiro.
“The goal is to mimic the same sorts of social interactions that the tests with the toys and balls measure, but without the toys and balls. The research has shown that the earlier autism can be spotted, the more beneficial intervention can be. And we want to provide everyone in the world with the ability to spot those signs as early as possible.”
The National Science Foundation, PhD scholarships from Brazil and the US Department of Defense, the Office of Naval Research, the National Geospatial-Intelligence Agency, the Army Research Office, and the Air Force Office of Scientific Research supported the research.
Source: Duke University