ADET MODEL: Real time autism detection via eye tracking model using retinal scan images

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Autism spectrum disorder (ASD) is characterized by concentration deficits with social stimuli.
• Early detection of autism is crucial and can be aided by analyzing visual attention patterns.
• Eye-tracking technology offers a potential method for identifying early autism biomarkers.

Purpose of the Study:

• To develop and evaluate an early autism detection model using eye-tracking data.
• To analyze children's visual behavior patterns through eye gaze sequences.
• To leverage Shi-Tomasi corner detection and vision transformers for autism diagnosis.

Main Methods:

• The proposed ADET model utilizes a corner detection-based vision transformer (CD-ViT).
• Input images are divided into patches and processed by a transformer encoder.
• The vision transformer is fine-tuned for binary classification using remora optimization on 547 eye-tracking images.

Main Results:

• The ADET framework achieved superior classification accuracy compared to RM3ASD, MLP, SVM, CNN, and other proposed methods.
• Specific accuracy improvements were noted, with the ADET framework demonstrating effectiveness in distinguishing between autistic and non-autistic children's visual patterns.

Conclusions:

• The developed screening method can assist medical professionals in accurate and efficient autism detection.
• Eye-tracking combined with advanced AI models like ADET provides a valuable tool for early identification of autism spectrum disorder.