1- Faculty of Psychology and Educational Sciences, Shahid Beheshti University, Tehran, Iran
2- Department of Psychology, Faculty of Psychology and Educational Sciences, Shahid Beheshti University, Tehran, Iran
3- Department of Electrical and Computer Engineering, Faculty of Electrical and Computer Engineering, University of Tehran, Iran |
Background and Purpose: Dyslexia is one of the most prevalent Specific Learning Disorders (SLD) and can adversely affect students’ academic performance and learning processes. So, the development of innovative approaches based on artificial intelligence technologies may facilitate the identification of distinctive eye-movement patterns for the early detection of dyslexia. Therefore, the present study aimed to differentiate eye-movement patterns during reading between students with dyslexia and typically developing students using artificial intelligence.
Method: The present study employed a cross-sectional design with a descriptive-analytical, causal-comparative method. The study population included all first to third-grade elementary school students in Tehran, Zahedan, Mashhad, and Qom city during the 2024–2025 academic year. The sample consisted of 586 students, including students diagnosed with dyslexia and typically developing students, recruited from comprehensive assessment, education, rehabilitation, and early developmental-educational intervention centers and general education schools in the aforementioned areas. Participants were selected through convenience sampling. Machine-learning models were employed to classify students based on eye-movement indices. The research instruments included the latest validated version of the Dyslexia Checklist (DC; Salimi Teymouri, 2007), the Intelligence Scale for Children-Fourth Edition (WISC-IV; Wechsler, 2003) in Iran, grade-appropriate reading texts for first- to third-grade students, and an eye-movement recording system. Data were analyzed using the Kruskal–Wallis test in SPSS - 27.
Results: The result showed significant differences between the typically developing students and students with dyslexia across all eye-movement measures (pupil movement: H = 43.09, gaze stability: H = 48.37, gaze stability ratio: H = 35.51, and not looking at the screen: H = 57.09; P < 0.001). For the selected models in the first, second, and third grades, accuracy was 0.90, 0.95, and 0.89, precision was 0.30, 0.80, and 0.33, sensitivity was 0.16, 0.40, and 0.26, and specificity was 0.98, 0.99, and 0.98, respectively. The area under the receiver operating characteristic curve was 0.95, 0.95, and 0.91, respectively. Gaze stability and pupil movement were also identified as the most important discriminative features across all three grades.
Conclusion: The findings show that eye-movement patterns can distinguish students with dyslexia from their typically developing peers with satisfactory accuracy. Accordingly, the integration of artificial intelligence–based screening tools, as a cost-effective and non-invasive approach for the early identification of dyslexia, is recommended as a strategic priority for educational authorities. |
