Live presentation

Asynchronous discussion


Combining eye tracking and machine learning techniques to detect dyslexia

Karina Viviana Rodriguez

  • Bahía Blanca,
  • Argentina
  • Karina Viviana Rodriguez ¹
  • , Francisco Iaconis ²
  • , Gustavo Gasaneo ¹
  • , Alejandra Mendivelzua ³
  • , Laura García Blanco ³
  • , Manuela Sanchez ³
  • 1 Departamento de Física UNS - IFISUR, Bahía Blanca - Centro Integral de Neurociencias Aplicadas, Bahía Blanca.
  • 2 Departamento de Física UNS - IFISUR, Bahía Blanca.
  • 3 Laboratorio Especializado en Aprendizaje y Neurociencias, Buenos Aires.

Recent studies estimate that 10 percent of the children population of Argentina suffers from Specific Learning Disorder. Dyslexia, being one of them, is considered a neurological learning disability [1]. In this presentation we combine machine learning techniques and eye tracking signals to explore the possibility of developing a tool to detect dyslexia in childrens.
The study performed included 25 neurotypical children evaluated at the school and a group of eleven children diagnosed with dyslexia evaluated in the psychopedagogue’s office. The task assigned to all of the subjects was to read on a monitor a short text designed for their age. During the reading the eye movements were recorded with an eye tracker mounted on the monitor.
In this work, the signals were processed using tools developed under python with the purpose of obtaining several variables such as mean time per fixation, number of saccades, mean size of saccades and other nine magnitudes. As part of the analysis performed, we applied Principal Component Analysis, Linear Discriminant Analysis and logistic regression to the data. The three techniques implemented were able to identify dyslexic from neurotypical children. We believe that the techniques implemented, combined with other tools, could help mental health professionals to diagnose.

[1] Asociación, AP. Manual diagnóstico y estadístico de trastornos mentales, Estados Unidos (2013).