A multidimensional eye-tracking assessment for estimating cognitive profiles in intellectual disability: A...

Area of Science:

• Neuroscience
• Developmental Psychology
• Computer Science

Background:

• Intellectual disability (ID) diagnosis requires resource-intensive expert assessments, limiting widespread use.
• Eye-tracking presents a potential digital biomarker for cognitive assessment, but its application to ID is limited.
• Current methods lack objective, accessible tools for estimating cognitive capacity in neurodevelopmental disorders.

Purpose of the Study:

• To develop and validate a novel eye-tracking assessment combined with deep learning for automated estimation of cognitive capacity in children with ID.
• To create a tool that can serve as a digital biomarker for neurodevelopmental disorders.
• To explore the potential of spatio-temporal gaze patterns in assessing cognitive subindices.

Main Methods:

• Developed three cognitive subtasks to elicit gaze patterns for verbal comprehension (VCI), fluid reasoning (FRI), and working memory (WMI).
• Collected data from children with ID and typically developing (TD) children.
• Compared a logistic regression (LR) model with predefined features against a convolutional neural network (CNN) trained on raw scanpath images.

Main Results:

• The CNN model achieved a superior 0.93 F1-score for subject classification, outperforming the LR model's 0.76 F1-score.
• CNN predictions from the working memory task correlated significantly with full-scale IQ and FRI/visuospatial (VSI) subscores.
• The deep learning model effectively captured higher-order reasoning and visuospatial processes.

Conclusions:

• Deep learning analysis of eye-tracking data offers a robust digital biomarker for cognitive capacity estimation.
• This approach paves the way for accessible and objective tools for assessing children with neurodevelopmental disorders.
• Eye-tracking combined with AI can provide valuable insights into cognitive profiles of ID.