Atypical postural control can be detected via computer vision analysis in toddlers with autism spectrum disorder

Géraldine Dawson, Kathleen Campbell, Jordan Hashemi, Steven Lippmann, Valerie A. Smith, Kimberly L. H. Carpenter, Helen L. Egger, Steven Espinosa, Saritha Vermeer, Jeffrey P. Baker, Guillermo SapiroScientific 2018 · DOI

Abstract

Evidence suggests that differences in motor function are an early feature of autism spectrum disorder (ASD). One aspect of motor ability that develops during childhood is postural control, reflected in the ability to maintain a steady head and body position without excessive sway. Observational studies have documented differences in postural control in older children with ASD. The present study used computer vision analysis to assess midline head postural control, as reflected in the rate of spontaneous head movements during states of active attention, in 104 toddlers between 16-31 months of age (Mean = 22 months), 22 of whom were diagnosed with ASD. Time-series data revealed robust group differences in the rate of head movements while the toddlers watched movies depicting social and nonsocial stimuli. Toddlers with ASD exhibited a significantly higher rate of head movement as compared to non-ASD toddlers, suggesting difficulties in maintaining midline position of the head while engaging attentional systems. The use of digital phenotyping approaches, such as computer vision analysis, to quantify variation in early motor behaviors will allow for more precise, objective, and quantitative characterization of early motor signatures and potentially provide new automated methods for early autism risk identification.

Introduction

While Autism Spectrum Disorder (ASD) is primarily diagnosed through social and communication challenges, motor differences often emerge much earlier. One such difference is postural control: the ability to maintain a steady head and body position. Historically, assessing these motor nuances required labor-intensive manual coding of video or expensive laboratory equipment like infrared motion trackers, making them difficult to scale for general pediatric screening. This study investigates whether a low-cost, automated system—using just a tablet’s front-facing camera—can detect subtle postural instabilities in toddlers during routine clinical visits.

Methods

The researchers developed a fully automatic Computer Vision Analysis (CVA) pipeline to quantify head movement in 104 toddlers (16–31 months old). The cohort included 22 toddlers diagnosed with ASD and 82 non-ASD toddlers (including typically developing children and those with other developmental delays).

Toddlers sat on a caregiver's lap and watched five minutes of engaging movies, such as "Bubbles," "Bunny," "Rhyme," and "Puppets." The CVA algorithm tracked 49 distinct facial landmarks at 30 frames per second to calculate 3D head pose angles (yaw, pitch, and roll). To ensure the measurement of involuntary "sway" rather than intentional gaze shifts, analysis was restricted to frames where the child was engaged (yaw < 20°). Head displacement was normalized by the child’s "eye width" to account for varying distances from the camera.

Facial landmark tracking (A) and experimental stimuli (B) used in the study.

Results

The study revealed a significant "stillness" deficit in toddlers with ASD. These children struggled to maintain a still, midline head position while their attentional systems were engaged. Toddlers with ASD exhibited a significantly higher rate of head movement across almost all movie stimuli compared to the non-ASD group.

The differences were most pronounced during complex, animated stimuli. For instance, during the "Bunny" movie, the rate of head movement for toddlers with ASD was 2.22 times higher than that of the non-ASD group (95% CI: 1.60, 3.07). This suggests that as the cognitive or sensory load of the stimulus increases, postural stability in toddlers with ASD decreases.

Time-series data showing head movement rates across different movie stimuli, comparing ASD (red) and non-ASD (blue) groups.

Discussion

This research demonstrates the potential of "digital phenotyping"—using sensors and algorithms to turn behavior into objective data. By moving from "top-down" human observation toward "bottom-up" data-driven biomarkers, clinicians can identify early motor signatures that are often too subtle for the naked eye.

While the results are promising, future work is needed to address the relatively small ASD sample size and to determine if these patterns are specific to ASD or shared with other neurodevelopmental conditions like ADHD. Integrating head movement data with other digital markers, such as eye tracking and facial expressions, could eventually create a highly accurate, automated risk profile for early ASD screening.

Summary updated

Abstract

Evidence suggests that differences in motor function are an early feature of autism spectrum disorder (ASD). One aspect of motor ability that develops during childhood is postural control, reflected in the ability to maintain a steady head and body position without excessive sway. Observational studies have documented differences in postural control in older children with ASD. The present study used computer vision analysis to assess midline head postural control, as reflected in the rate of spontaneous head movements during states of active attention, in 104 toddlers between 16-31 months of age (Mean = 22 months), 22 of whom were diagnosed with ASD. Time-series data revealed robust group differences in the rate of head movements while the toddlers watched movies depicting social and nonsocial stimuli. Toddlers with ASD exhibited a significantly higher rate of head movement as compared to non-ASD toddlers, suggesting difficulties in maintaining midline position of the head while engaging attentional systems. The use of digital phenotyping approaches, such as computer vision analysis, to quantify variation in early motor behaviors will allow for more precise, objective, and quantitative characterization of early motor signatures and potentially provide new automated methods for early autism risk identification.
BibTeX
@inproceedings{dawson2018atypical,
  author = {Géraldine Dawson and Kathleen Campbell and Jordan Hashemi and Steven Lippmann and Valerie A. Smith and Kimberly L. H. Carpenter and Helen L. Egger and Steven Espinosa and Saritha Vermeer and Jeffrey P. Baker and Guillermo Sapiro},
  title = {{Atypical postural control can be detected via computer vision analysis in toddlers with autism spectrum disorder}},
  booktitle = {Scientific},
  year = {2018},
  doi = {10.1038/s41598-018-35215-8},
}
Figure