Research demonstrates that early interventions are critical for overcoming the social and behavioral deficits of autism spectrum disorders (ASD). That is why it is important for clinicians to be able to identify individuals with autism as early as possible. Unfortunately for many, autism assessments may not be available for various reasons. To solve this problem, researchers are seeking new ways to streamline autism assessments so that anyone can access autism-related services.
Two studies this year investigated how video can be a part of testing for ASD. One study, “The potential of accelerating early detection of autism through content analysis of YouTube videos,” published in PLOS One, involved assessing YouTube videos to determine if the video’s subject demonstrated autism-related behaviors. A second study, “Computer vision tools for low-cost and noninvasive measurement of autism-related behaviors in infants,” published in Autism Research and Treatment, described the creation of a computer algorithm that can detect when an infant engages his or her attention. These studies demonstrate that video can be used effectively and inexpensively to identify autism-related behaviors.
A third study found a way to make autism diagnosis simpler. In “Validation of the modified checklist for autism in toddlers, revised with follow-up,” published in Pediatrics, researchers described updating the Modified Checklist for Autism in Toddlers (M-CHAT). The researchers updated the M-CHAT to offer a faster assessment in a way that more people can understand. Like the video-based studies, this work involved a method that made autism diagnosis more accessible.
Video Screening for Autism
Approximately 27 percent of eight-year-olds with autism are undiagnosed in the United States. Although most children with autism receive a diagnosis around age four, there are many who do not. As children age, it becomes more difficult to assess their early years, which can make diagnosing autism more challenging. Parents and caregivers may not have completely reliable memories of early childhood development, or they may not have noticed any of the characteristics of autism spectrum disorders (ASD).
That is why a team from the Center for Biomedical Informatics at Harvard Medical School tested a way to evaluate people for ASD through home videos. Using videos available on YouTube, the researchers evaluated video subjects for autism. They selected YouTube videos related to autism and to children, searching with keywords like ASD, autistic behavior, and birthday party. They only chose videos that were less than 10 minutes long, had clear audio quality, and focused on the activities of a child between age 1 and 15.
To evaluate the videos, the researchers used module one of the Autism Diagnostic Observation Schedule (ADOS). Module one primarily focuses on the behavioral characteristics of younger children. Because younger children can benefit most from early diagnosis, the research team elected to use ADOS module one.
Non-clinical raters scored the videos—45 of children with an ASD and 55 of children without an ASD. The raters accurately identified 98.6 percent of the videos as representing autism or not. One trained clinician rated 22 of the videos. The non-clinical raters and the trained clinician’s scores agreed on 19 of the 22 videos.
The findings indicate that it may be possible to diagnose—or at least screen for—autism using home videos. Even though the videos depicted disparate conditions and were not controlled like a clinical environment, the raters were able to accurately indicate whether a video’s subject was demonstrating autism-related behaviors. The authors state that, “The high classification accuracy (98.6%) supported the feasibility of applying ADOS module 1 questions to videos on YouTube. In addition, the high sensitivity (94.1%) and perfect specificity suggested that home video classification could yield both high positive and negative predictive value” (Fusaro, et. al., 2014).
This study could help expand the ability to diagnose children early, thus providing children with early interventions. Using video screenings could also promote faster screening times and provide access to clinical services for people in remote areas.
Video Assessment of Autism Through Movement
Another study examined how video could be used to support autism diagnosis. Instead of relying on previously recorded videos, the research team at Duke University and the University of Minnesota recorded videos in a clinical environment and created a movement-analyzing algorithm to assess infants. The researchers wanted to create a method that could compete with a trained reviewer in a nonintrusive way. Because assessments of autism in children are based on movement, the research team considered a nonintrusive approach to be one that does not restrict movement in any way.
The researchers conducted their evaluations using two behavior types from the Autism Observation Scale for Infants (AOSI): disengagement of attention and visual tracking. They used computer vision algorithms for assessing the infant’s visual attention, tracking head movement on the yaw (left-right) and pitch (up-down) axes. The system tracks the infant’s left ear, left eye, and nose and generates an estimate based on their position and how long these features remain in one position. The computer vision algorithms can automatically detect when an infant disengages from a task.
The participants were 12 infants aged 5 to 18 months, some male and some female. The infants were considered at-risk for autism based on being the baby sibling of a child with autism, exhibiting developmental delays, or being born prematurely. A clinician performed two tests with the infants. An inexpensive camera placed two feet away from the infant recorded the procedure.
To test disengagement of attention, defined as the “ability to disengage and move eyes/attention from one of two competing visual stimuli,” the clinician shook a noisy toy. After the infant engaged his or her attention with the toy, the clinician shook a second toy on the opposite side of the infant. The clinician shook the toys until the infant shifted attention. A delayed response—a one to two second lag—is associated with autism.
To test visual tracking, defined as the “ability to visually follow a moving object laterally across the midline,” the clinician moved a rattle silently at eye level from one side of the infant to the other. Gaze tracking that is interrupted, delayed, or partial is associated with autism.
The clinician and the computer vision algorithm agreed about the infants’ performance in the majority of trials. For the attention disengagement task, the clinician rated 24 as “pass” and three as “delayed.” The algorithm agreed with 23 of the clinician’s 24 passes, scoring the other as a delay. Of the three delays, the algorithm agreed with two. For the visual tracking task, the clinician scored 14 as “pass,” of which the algorithm agreed with 13. The algorithm scored the fourteenth as “interrupted.” The clinician scored two other trials as “partial.” The algorithm agreed with one partial assessment, but scored the other as “interrupted.”
In addition to the algorithm and the clinician, the researchers asked two non-expert raters, both psychology students, to assess the trials. The non-experts agreed with the clinician on approximately half of the attention disengagement trials and around 60 percent of the visual tracking trials. In contrast, the algorithm agreed with the clinician 86 percent of the time.
If this method is validated, “it would provide accurate quantitative measurements for assessing infant visual attention … improving the shareability of clinical records” (Hashemi, et. al., 2014). This method could result in objective, accurate measures for evaluating infants for autism spectrum disorders. Using their method is also relatively low cost and could expand access to autism services.
Streamlining a Popular Evaluation of Autism
Researchers are finding ways to streamline autism’s diagnostic process without video, too. A group of researchers from Georgia State University tested an updated version of the Modified Checklist for Autism in Toddlers. The M-CHAT is a popular tool for diagnosing young children that was originally developed in the early 1990s.
To streamline the M-CHAT, the researchers simplified the test’s language and scoring mechanisms. They rephrased questions to improve understandability for people with limited education. For example, questions that start with “Does your child take an interest … ” became “Is your child interested … ” The researchers illustrated questions with examples that were more age-appropriate. In the original test, one question asks if children solicit others to play on the playground. However, that behavior is something parents are more likely to observe in four-year-olds, not the 16- to 3-month-olds the M-CHAT evaluates.
Scoring the updated M-CHAT is simple. If a child’s caregiver answers “yes” to three or more of the test’s initial 20-question screening, the child is immediately referred for a more thorough evaluation. All of the questions are weighted equally to streamline the process.
The researchers tested the updated M-CHAT with 16,071 toddlers in Georgia and Connecticut during routine 18- and 24-month doctor visits. The toddlers who scored at least three on the parent questionnaire and at least two on the follow-up interview have a 50 percent chance of being diagnosed with an ASD. They also had a 95 percent chance of being diagnosed with some form of developmental delay.
The updated M-CHAT decreases the number of false positives identified using the original M-CHAT, but it may also leave behind some children who do have autism. The researchers identified 18 children with autism who did not score high enough on the initial screening to warrant a follow-up evaluation.
The updated screening suggests an autism prevalence of 0.67 percent, compared to the original test’s 0.45 percent. Both projections are lower than the 1.1 percent prevalence of autism in the United States today.
The updated M-CHAT has the potential to simplify the process of receiving an autism diagnosis. The new test allows for clinicians to complete an assessment in one day, eliminating the need for return visits. This could help children get access to interventions sooner.
Autism Assessments in Your Practice
Researchers are working on novel methods of evaluating autism so that clinicians can identify autism spectrum disorders in patients faster and earlier. The sooner patients can be identified as having autism, the quicker they are able to receive critical interventions. The earlier an intervention, the more effective it is.
Certified Autism Specialists are committed to using up-to-date, peer-reviewed methodologies as they find ways to support patients with unique needs. Finding new ways to expand your practice and evaluate patients for autism helps you connect patients with the services they need to improve their lives.
Fusaro, V. A., Daniels, J. Duda, M., DeLuca, T., D’Angelo, Tamburello, J. … Wall, D. P. (2014). The potential of accelerating early detection of autism through content analysis of YouTube videos. PLOS One, 9(4). doi: 10.1371/journal.pone.0093533
Hashemi, J., Mariano, T., Spina, T. V., Esler, A., Morellas, V., Papanikolopoulos, N., … Sapiro, G. (2014). Computer vision tools for low-cost and noninvasive measurement of autism-related behaviors in infants. Autism Research and Treatment, 2014, 1-12. doi:10.1155/2014/935686
Robins, D. L., Casagrande, K., Barton, M., Chen, C. M., Dumont-Mathier, T., & Fein, D. (2014). Validation of the modified checklist for autism in toddlers, revised with follow-up. Pediatrics 133(1), 37-45. doi: 10.1542/peds.2013-1813