Multimodal AI Framework for Early Detection of Dyslexia in Children

Abstract

Dyslexia is a lifelong learning disorder. It is associated with how the brain functions, making it difficult for children to read, write, and retain information. If it is not identified when a child is young, then it can make it very challenging for the child to perform well at school. It can also negatively affect the child’s self-esteem. Therefore, it is very important to identify the problem during early childhood. This is even more critical in a country such as Pakistan, where awareness of dyslexia is limited and early screening facilities for children are not widely available. Usually, Traditional methods used to determine whether children have dyslexia, the process is slow, may produce inaccurate conclusions, and often examines only a single aspect of the disorder. To address these limitations, this document proposes an approach in which multiple features are integrated into a computer-based system to identify dyslexia. The program analyzes handwriting patterns and cognitive processing characteristics. When evaluated using data from data 200 children, the program achieved 96.9% accuracy and 91.7% sensitivity in determining whether a child has dyslexia. It successfully identified all cases, including the small number of children diagnosed with dyslexia, while only five children without dyslexia were incorrectly classified. Although this study has certain limitations, including a relatively small dataset and related constraints, the findings indicate that this multimodal behavioral and perceptual analysis approach improves the identification of children with dyslexia. The proposed approach offers a practical solution for early screening in educational settings, with potential for further improvement through additional data modalities.