Decoding Dyslexia Through EOG Complexity: A Machine Learning Driven Analysis of Reading Behavior in Children

Dyslexia is one of the most prevalent learning disabilities and can be briefly called a special reading disability. Eye movement analysis is an important part of the reading evaluation process. Rapid recognition of dyslexia is critical in terms of overcoming learning difficulties by planning education accordingly. This paper presents a human–computer interface system that analyzes eye movements to determine the font type that may help children with dyslexia improve their reading skills. In this study, various features were extracted by performing a nonlinear examination of Electrooculogram (EOG) signals recorded from children with dyslexia and healthy (control) children. These features were obtained from complexity measurements using several entropy algorithms, Lempel-Ziv complexity (LZC), and Kolmogorov complexity (KC). The success of the features in differentiating the EOG signals of children with dyslexia was measured using classification with different machine learning algorithms. The statistical analysis results showed that the features extracted from the EOG signals of dyslexia and control groups were significantly different between the groups. With the Decision Tree (DT) algorithm, an accuracy of 84.4% was obtained, and with the Support Vector Machine (SVM), the area under the ROC curve was 89.9%. According to our research, EOG complexity features were sufficient to detect physiological changes in both healthy children and dyslexia patients. To the best of our knowledge, this study is the first to conduct a complexity analysis of EOG-based dyslexia detection, yielding the results obtained.