Early autism detection is crucial for effective interventions. Current methods, such as caregiver questionnaires like the MCHAT-R/F, have limitations in accuracy and may be influenced by socioeconomic factors. Motor impairments are frequently observed in autistic children, often appearing early and potentially serving as an early indicator. Existing assessment methods for motor skills lack scalability and objectivity. This study aims to address this gap by developing and validating a scalable, quantitative, and engaging tablet-based assessment of visual-motor abilities in young autistic children. The assessment is based on a bubble-popping game that leverages computer vision to objectively measure various touch-based features. The study will also investigate the impact of co-occurring ADHD on motor function and correlate game-based measures with standardized assessments of cognitive and motor abilities. The researchers hypothesized that autistic children would exhibit distinct performance patterns on the bubble-popping game compared to neurotypical children, with further differentiation based on co-occurring ADHD. The use of a gamified approach on readily available tablets offers a potentially efficient and widely applicable method for early screening and assessment.

Extensive research demonstrates a strong correlation between early motor impairments and autism spectrum disorder (ASD). Prevalence estimates of motor impairments in autism range from 50-85%, although this may be an underestimation due to limitations of current assessment tools. Motor difficulties are often among the earliest reported signs of autism and are present even in those without cognitive impairment. Various aspects of motor skills have been investigated in ASD, including gait, balance, coordination, accuracy, reaction time, and dexterity, using diverse methodologies. Studies on infants at high risk for ASD have shown that early motor skill development is a predictor of later language acquisition and ASD diagnosis, highlighting the importance of early assessment. Challenges in visual-motor integration are common in ASD, potentially impacting social skill development due to limitations in imitation. The development of miniaturized sensors and mobile devices has facilitated the use of tablet-based assessments and games for evaluating motor skills, producing large datasets suitable for quantitative analysis. Previous studies using these technologies have shown promise in identifying motor signatures in ASD and predicting disease severity in other neurological conditions. This work extends prior research by assessing a gamified visual motor assessment in toddlers, controlling for the effects of cognitive ability and co-occurring ADHD, and exploring the relationships between novel visual-motor features and clinical profiles.

This study used a two-study design. Study 1 included 151 children (18-36 months) from four Duke pediatric primary care clinics, with 23 having ASD. An age-matched neurotypical group was also selected. Study 2 involved an independent sample of 82 children (36-120 months), including 63 with ASD (32 with co-occurring ADHD) and 19 neurotypical children. Participants were excluded for motor impairments, language barriers, caregiver interference, and insufficient engagement. Informed consent was obtained from caregivers, and the study was approved by the IRB. In Study 1, caregivers completed the M-CHAT-R/F; those who failed were referred for further diagnostic evaluation using ADOS-2 and Mullen Scales of Early Learning (MSEL). Study 2 participants received ASD diagnosis via ADOS-2 and ADI-R, ADHD diagnosis (when applicable) using MINI-Kid, ADHD-RS, and clinical judgment, and cognitive assessment using the Differential Abilities Scale (DAS). The core assessment involved a bubble-popping game on a tablet. The game involved popping bubbles appearing in vertical tracks on the screen. Computer vision analyzed various touch-based features, including number of touches, number of pops, popping rate, double touch rate, screen exploratory percentage, number of targeted bubbles, number of transitions, repeat percentage, touch duration, touch length, touch velocity, applied force, distance to center, popping accuracy, and time spent on a target. Statistical analysis included proportion Z-tests for game experience, Mann-Whitney-U tests for age and IQ comparisons, ANCOVA for group comparisons of motor variables (covariates: age and IQ), Spearman's rho correlation for relationships between motor features and clinical measures, and logistic regression with leave-one-out cross-validation to evaluate model performance and the combined power of the features in predicting diagnostic groups. The ROC curves and AUC were utilized for evaluating the performance of the models. Benjamini-Hochberg correction controlled for FDR.

Younger (1.5-3 years) autistic children demonstrated a lower bubble popping rate and less accurate center touches than their neurotypical counterparts, along with longer touch durations and greater variability in performance. The number of touches was similar across groups, indicating similar levels of engagement. Older (3-10 years) autistic children showed a lower average touch frequency and shorter median time spent targeting bubbles than neurotypical children. In older children, the presence of co-occurring ADHD was associated with lower accuracy (larger distance to the center), lower bubble popping rate, a higher number of touches per target, and increased variability in motor behavior. Combining multiple game features improved group discrimination in both age groups. Logistic regression models incorporating multiple features showed AUC values of 0.73 (95% CI, 0.63-0.83) for distinguishing younger autistic and neurotypical children and 0.74 (95% CI, 0.62-0.86) for differentiating older autistic children with and without ADHD. For Study 1, the fine motor T-score on the MSEL correlated positively with popping rate, double touch rate, and average popping accuracy, and negatively with average touch velocity, average and standard deviation of touch duration, and variability of the maximum popping accuracy. The early learning composite score positively correlated with the number of pops and average popping accuracy. Expressive and receptive language scores correlated with screen exploration and repetitive behavior. In Study 2, IQ correlated with the number of pops, while spatial skills were associated with stronger visual-motor skills (more pops, shorter touch duration and velocity, lower force variation). Verbal skills showed only a weak correlation with the number of touches.

The findings support the use of the tablet-based bubble-popping game as a feasible and effective tool for assessing visual-motor skills in children with autism. The game is engaging, easily administered, and objectively quantifies multiple aspects of motor performance, including accuracy, variability, and exploratory behavior. The observed differences in motor skills between autistic and neurotypical children, particularly in younger children, highlight the potential for using this game as a component of broader autism screening tools. The results underscore the importance of considering co-occurring conditions, such as ADHD, when evaluating motor skills in autistic children. The correlations between game-based features and standardized assessments demonstrate the validity of the measures obtained from the game. Future research should focus on longitudinal studies to track the development of visual-motor skills in autistic children and the impact of interventions. The inclusion of additional features, such as gaze and facial expressions, might further enhance the diagnostic accuracy of this assessment.

This study demonstrates the potential of a simple, engaging, tablet-based bubble-popping game as a scalable tool for assessing visual-motor skills in children with autism. The game effectively captures subtle differences in motor performance between autistic and neurotypical children, as well as the impact of co-occurring ADHD. The findings highlight the feasibility of using this approach for early identification of autism-related motor impairments and suggest promising avenues for future research focused on longitudinal studies and the integration of other behavioral measures. Further refinement of the game and its features, alongside larger-scale validation studies, could lead to a valuable clinical tool supporting early diagnosis and intervention for autism.

This study's limitations include the relatively small sample size, particularly for subgroup analyses based on sex and ethnicity, limiting the generalizability of the findings. Differences in clinical assessments across the two studies hindered broader comparisons of motor performance and clinical variables. The game's duration (20 seconds) might not be sufficient for assessing learning, focus, and anticipation. In Study 1, the lack of comprehensive diagnostic and cognitive evaluations for the neurotypical group is a limitation. The absence of assessment for comorbidities other than ADHD might also affect the findings.