Tablet-based gameplay identifies movement patterns related to autism spectrum disorder

Anzulewicz, Anna and Sobota, Krzysztof and Delafield-Butt, Jonathan (2018) Tablet-based gameplay identifies movement patterns related to autism spectrum disorder. In: Cognitive Neuroscience Society 25th Annual Meeting, 2018-03-24 - 2018-03-27, Sheraton Hotel.

[thumbnail of Anzulewicz-etal-CNS-2018-Tablet-based-gameplay-identifies-movement-patterns-related-to-autism-spectrum-disorder]
Preview
Text. Filename: Anzulewicz_etal_CNS_2018_Tablet_based_gameplay_identifies_movement_patterns_related_to_autism_spectrum_disorder.pdf
Accepted Author Manuscript

Download (586kB)| Preview

Abstract

Background: It has been proposed that one of the early markers of autism spectrum disorder (ASD) are abnormalities in the development of intentional movements, which can be observed from early childhood. New evidence suggests that disruption of motor timing and integration may underpin the disorder, providing a new potential marker for its identification. Objectives: In this study, we used widely available tablet devices (iPads) to identify differences in kinematics between children diagnosed with ASD and their typically developing (TD) peers. We also compared movement patterns of children diagnosed with neurodevelopmental disorders other than autism (OND) with movement patterns exhibited by ASD and TD children. We utilised tablet devices’ inertial sensors (accelerometer, gyroscope, and touchscreen to record the movements children make while playing two educational games on a tablet. Methods: Ninety-six children (aged 3-6) diagnosed with ASD, 37 diagnosed with OND, and 387 TD children took part in the study. The children were asked to play two educational games on a tablet. Each game consisted of two parts: two-minute long training and five-minute long test session. During the gameplay, we collected data from tablet’s sensors and screen. After the experimental session, 262 variables obtained by simple calculation of the raw sensor data (e.g. acceleration of the movements) were extracted and analysed using machine learning algorithms. To increase generalisation properties of the models, we reduced dimensionality to 49 most significant variables. Results: To compare movement patterns of children with ASD, OND, and TD children, we used machine learning algorithms. Each algorithm differentiated individuals within the ASD group from the other groups using 49 variables derived from the touch screen and inertial sensors. ASD - TD comparison: The algorithms classified children diagnosed with ASD from TD children with up to 93% accuracy. OND - TD comparison: The algorithms classified children diagnosed with OND from TD children with up to 95% accuracy. The results suggest that movement patterns of typically developing children are different than patterns exhibited by children diagnosed with neurodevelopmental disorders other than autism. ASD - OND comparison The algorithms classified children diagnosed with ASD from OND children with up to 93% accuracy. This result suggests that ASD is characterised by movement patterns that can be differentiated from patterns related to other neurodevelopment disorders. Conclusions: These findings support the view that children with ASD can be differentiated from TD children by movement patterns analysis. We also provide evidence suggesting that patterns characteristic of ASD children are different from patterns exhibited by children with OND. However, the latter result is not particularly strong due to the small sample of OND group. Further research is needed to provide better evidence.