Defining the ‘Autism Motor Signature’: Characterisation of Motor Patterns of Children with Autism during Ipad Gameplay

Background:

New evidence indicates disruption to motor timing and integration may underpin autism spectrum disorder (Trevarthen & Delafield-Butt, 2013). In previous work, we employed smart tablet computers with touch-sensitive screens and embedded inertial movement sensors to record the movement kinematics and gesture forces made by children with autism and children developing typically. Machine learning analysis of those children’s motor patterns identified autism with 93% accuracy, supporting the notion disruption to movement is core feature of autism, and demonstrating autism can be computationally assessed by fun, smart device gameplay (Anzulewicz, Sobota, & Delafield-Butt, 2016). Machine learning analysis is a powerful tool, but it is a ‘black box’ approach. Here, we present new analyses to better define the nature of the motor disruption and therefore to better characterize the ‘autism motor signature’.

Objectives:

(1) Determine the kinematic features that characterize the autism motor signature during iPad gameplay.

(2) Define these in terms of motor control properties to better understand its nature.

Methods:

37 children 3-6 years old clinically diagnosed with Childhood Autism and 45 age- gender-matched children developing typically were included. Children with sensory or motor impairment were excluded. iPad mini tablet computers employed two education games: (1) ‘Sharing’ where the main gameplay was to divide a piece of food (e.g. an apple) and distribute it evenly among four children on the screen; and (2) Creativity where gameplay was open, unstructured colouring of a toy or animal. Children were seated at a table and the iPad placed in front of them. A five minute unassisted ‘experimental’ phase was preceded by a two minute assisted ‘training’ phase. Data was collected during ‘experimental’ gameplay from the tablets' inertial sensors (tri-axial accelerometer, tri-axial gyroscope and magnetometer) and touch screen. Raw values and simple kinematic calculations (e.g. gesture duration, amplitude, acceleration, etc.) produced 262 ‘features’ analysed in previous machine learning work with a classification accuracy of 93%. These same features were extracted and analysed using standard statistical multivariate analyses.

Results:

Features associated with greater forces made at contact and forces put into the device within a gesture were greater and more variable in children with autism than in children developing typically, respectively. Further, gesture kinematics were faster and larger, with more distal use of space in children with autism. And children with autism made faster taps on the screen.

Conclusions:

iPad gameplay stands in agreement to standard measures, providing and accessible new paradigm for autism research. A particular motor pattern characterised by greater impact forces was apparent, supporting the notion that termination of a movement is different, difficult, or disrupted in autism. This finding corroborates data in reach-to-grasp and reach-to-place paradigms that demonstrate greater velocity at the termination of a movement (Crippa et al., 2015). Identification of greater variation in the distribution of forces during a gesture corroborates with reach-to-grasp motion capture data (Cook et al., 2013; Torres et al., 2013). Thus, it appears sub-second regulation of children’s intentional movements and timing their termination are two key features of the autism motor signature.