32038
Drawing Patterns during a Smart Tablet Colouring Game: A New Analysis to Identify Autism Spectrum Disorders in Early Childhood.
Evidence suggests the prospective motor organisation of intentional movements is be disrupted in Autism Spectrum Disorders (ASD) (Trevarthen and Delafield-Butt, 2013). Recently, machine learning analysis of children’s motor patterns made during smart tablet gameplay was found to differentiate between children Typically Developing (TD) from those developing with ASD with 93% accuracy (Anzulewicz, Sobota, and Delafield-Butt, 2016). Here, we employed the same serious games in a new study to assess play patterns in the free-style colouring game, ‘Creativity’, which performed with greater predictive accuracy. We analysed children’s motor patterns during colouring, implementing an observational approach.
Objectives:
A description of ASD colouring behaviours using a colouring serious game on an iPad.
Methods:
Participants. 70 children, 2 to 6 years-old (40=TD; 30=ASD). TD recruited from nurseries in Glasgow, UK; 24 ASD recruited through the Scottish Centre for Autism (NHS Greater Glasgow and Clyde), and 6 of collected at the Gillberg Neuropsychiatry Centre, Gothenburg, Sweden.
Procedure. The children were seated at a table with the iPad resting in front of them. The iPad was equipped with 2 game apps: (a) Sharing, that involved dividing food among 4 cartoon characters; and (b) Creativity, free-style colouring of chosen pictures. They played with each game for 5 minutes, plus 2 minutes for pre-test familiarisation. In Creativity, the children could choose to draw and colour one image or more during the play. Thus, Creativity was divided into two phases: (i) Tracing the chosen image and (ii) Colouring the chosen image.
Data Analysis. The number of image chosen, overall time spent in the Tracing and Colouring phases, and the amount of screen area touched were calculated. Subgroups were used to calculate Time spent in Tracing and Colouring phases per image as well as its Tracing accuracy.
Results: The Tracing phase duration appeared comparable between the two groups for each image, although ASD children spent cumulatively more time in this phase, because children with ASD generally made more drawings. Conversely, children with ASD spent less time in the Colouring phase compared to their peers. Further, they showed higher variance in Tracing accuracy. Interestingly, 6 children with ASD created “non-drawings”, which were figures that did not match the set of colouring templates, and some adapted drawings into “non-drawing”.
Conclusions:
Differences in the drawing patterns of children with ASD during the colouring game could be interpreted as a different approach to play that with TD children. ASD children would make their own scribbles, as “non-drawings”, and use the Colouring phase as a “free drawing” task. The significant reduction of duration in Colouring in ASD children may indicate these children were “locked” in a loop of Tracing and/or picture choosing, comparable to repetitive behaviours. In sum, these behavioural features may provide new variables to be included in machine learning data analytics in future work to improve its predictive accuracy in identification of autism during early childhood.