Towards Autism Screening through Emotion-guided Eye Gaze Response.

Area of Science:

• Neuroscience
• Computer Science
• Developmental Psychology

Background:

• Autism Spectrum Disorder (ASD) is characterized by social interaction deficits.
• Non-verbal communication, including eye gaze and facial expressions, is often atypical in ASD.
• Pupil response has shown potential for ASD screening, but the influence of emotional stimuli is under-explored.

Purpose of the Study:

• To design, develop, and evaluate a lightweight LSTM model for ASD detection using pupil responses.
• To investigate the impact of emotional stimuli (angry, happy, neutral) on pupil responses in individuals with and without ASD.
• To assess the efficacy of pupil response analysis for screening ASD during short social interactions.

Main Methods:

• A lightweight Long Short-Term Memory (LSTM) model was developed to analyze pupil responses.
• Pupil data, including diameter, fixation duration, and location, were captured during interactions with a virtual agent.
• The model was trained and evaluated on detecting ASD sessions based on these pupil response metrics.

Main Results:

• Significant variations in pupil responses were observed in ASD sessions compared to non-ASD sessions.
• The LSTM model achieved an average accuracy of 77% in detecting ASD.
• Accuracy increased to over 80% when utilizing pupil responses to angry and happy emotional stimuli.

Conclusions:

• Pupil response analysis, particularly when influenced by emotional stimuli, is a promising method for ASD screening.
• The developed lightweight LSTM model demonstrates effective ASD detection capabilities.
• Further research into emotional influences on pupil dynamics can enhance ASD diagnostic tools.