Accurate and Early Diagnosis of Children at Risk for Dyslexia: Comparison of Two Intelligent Systems Designed by Artificial Neural Network

The aim of the present study was early and accurate diagnosis of preschoolers at risk for dyslexia through designing an intelligent diagnostic system.

The current research was a “research and development” type of investigation, in terms of its goal, and a descriptive research, assessment, and diagnostic type study, in terms of data collection method. The Neuro-cognitive program designed by Delavarian et al. was used for evaluation of the children. The efficacy, accuracy, validity, and reliability of this program were proven in many previous studies. Participants were selected following cluster random sampling method and their neuro-cognitive functions were saved for two years until the definite diagnosis of each individual was determined and then the data was applied in designing the diagnostic system. Multilayer perceptron and radial basis function artificial neural networks were applied in designing the system and they were compared according to their accuracy and sensitivity.

The average accuracy of the system in early diagnosis of preschoolers at risk for dyslexia, designed by multilayer perceptron neural network, reached to 94.40% and the network’s sensitivity and specificity were obtained to be 90.27 and 95.28%, respectively.

According to the high validity and reliability of the neuro-cognitive program and the high accuracy and sensitivity of the designed decision support system, the mentioned system could be used in early detection of at risk preschoolers, prior to entering the elementary school.