Predictive Model for Dyslexia from Fixations and Saccadic Eye Movement Events.

Area of Science:

• Neuroscience
• Cognitive Science
• Computer Science

Background:

• Dyslexia is a reading disorder impacting phonological processing and word decoding, not a visual impairment.
• Difficulties in dyslexia manifest as altered eye movement patterns during reading.
• Eye-tracking is a valuable tool for analyzing cognitive processing and identifying reading disorders.

Purpose of the Study:

• To identify individuals with dyslexia using eye-tracking data.
• To determine key eye movement features that differentiate dyslexics from non-dyslexics.
• To evaluate the effectiveness of machine learning models in classifying dyslexia.

Main Methods:

• Eye movement features (fixations, saccades) were extracted using statistical, dispersion (I-DT), and velocity (I-VT) threshold algorithms.
• Machine learning models, including Support Vector Machine (SVM), Random Forest (RF), Logistic Regression (LR), K-Nearest Neighbor (KNN), and a Hybrid SVM-PSO model, were employed for classification.
• Feature selection focused on metrics like fixation duration, saccade duration, and movement counts.

Main Results:

• The Hybrid SVM-PSO model achieved the highest classification accuracy at 95.6%.
• Key features contributing to accurate classification included average fixation count, average fixation duration, average saccadic duration, and total saccadic movements.
• Velocity-based algorithms for feature detection outperformed dispersion-based algorithms and statistical measures.

Conclusions:

• Eye movement analysis using machine learning offers a promising approach for dyslexia identification.
• Velocity-based feature extraction methods are more effective for distinguishing dyslexic reading patterns.
• Specific eye movement metrics provide reliable indicators for differentiating dyslexic and non-dyslexic readers.