Automatic Autism Spectrum Disorder Detection Thanks to Eye-Tracking and Neural Network-Based Approach

Romuald Carette (1) , Federica Cilia (2) , Gilles Dequen (1) , Jerome Bosche (1) , Jean-Luc Guerin , Luc Vandromme (3)

1 MIS - Modélisation, Information et Systèmes - UR UPJV 4290
2 CRP-CPO - Centre de Recherche en Psychologie : Cognition, Psychisme et Organisations - UR UPJV 7273
3 CHIMERE - CHirurgie, IMagerie et REgénération tissulaire de l’extrémité céphalique - Caractérisation morphologique et fonctionnelle - UR UPJV 7516

Autism spectrum disorder (ASD) is a neurodevelopmental disorder quite wide and its numerous variations render diagnosis hard. Some works have proven that children suffering from autism have trouble keeping their attention and tend to have a less focused sight. On top of that, eye-tracking systems enable the recording of precise eye focus on a screen. This paper deals with automatic detection of autism spectrum disorder thanks to eye-tracked data and an original Machine Learning approach. Focusing on data that describes the saccades of the patient's sight, we distinguish, out of our six test patients, young autistic individuals from those with no problems in 83% (five) of tested patients, with a results confidence up to 95%.

Domaines: Sciences de l'Homme et Société

Romuald Carette, Federica Cilia, Gilles Dequen, Jerome Bosche, Jean-Luc Guerin, et al.. Automatic Autism Spectrum Disorder Detection Thanks to Eye-Tracking and Neural Network-Based Approach. INTERNET OF THINGS (IOT) TECHNOLOGIES FOR HEALTHCARE, HEALTHYIOT 2017, 2017, Angers, France. pp.75-81, ⟨10.1007/978-3-319-76213-5\_11⟩. ⟨hal-03599565⟩