Classifying Individuals with ASD Through Facial Emotion Recognition and Eye-Tracking

Area of Science:

• Neuroscience
• Psychology
• Computer Science

Background:

• Individuals with Autism Spectrum Disorder (ASD) exhibit atypical visual scanning patterns during face and emotion perception.
• Previous research has utilized eye-tracking data to characterize ASD, but integration with task performance for diagnostic purposes remains an area for exploration.

Purpose of the Study:

• To investigate the utility of combining eye movement data with facial emotion recognition task performance for identifying individuals with ASD.
• To develop and evaluate a machine learning model for distinguishing between individuals with ASD and typically developing (TD) controls.

Main Methods:

• A cohort of 23 individuals with ASD and 35 TD controls completed the Dynamic Affect Recognition Evaluation (DARE) task, involving emotion recognition from facial videos.
• Eye movements, response times, and recognition accuracy were recorded during the DARE task.
• A machine learning approach was developed, integrating task performance, gaze patterns, and deep neural network-extracted facial features to classify participants.

Main Results:

• Significant differences were observed in response times and eye movement patterns between individuals with ASD and TD controls.
• Recognition accuracy did not significantly differ between the two groups.
• The proposed machine learning model achieved an 86% classification accuracy, demonstrating comparable performance to standardized diagnostic scales.

Conclusions:

• Combining eye movement data and task performance offers an efficient and objective method for identifying individuals with ASD.
• The machine learning model highlights the potential of integrating behavioral and visual data for ASD assessment.
• Further analysis of distinguishing features can elucidate social and attentional deficits associated with ASD.