Learning disabilities such as dyslexia, dyscalculia, and dysgraphia tend to remain undiagnosed during early childhood owing to a lack of standardized assessments, requirement of specialized professionals to conduct them, and the complexity involved in diagnosis. This research focuses on designing an early screening tool with the help of Artificial Intelligence and a mobile application that aims to serve as the initial marker for assessing whether there are any issues related to reading skills, mathematical abilities, or handwriting among 5-7-year-old children. There are three main modules included in the proposed system: (a) arithmetic module that relies on quizzes to identify the challenges associated with dyscalculia; (b) handwriting analysis module with deep learning algorithm and a VGG16-based convolutional neural network to detect possible risks of dysgraphia; and (c) reading module that utilizes games to analyze phonological problems related to dyslexia. The data consists of 395 handwritten samples obtained from several educational institutions supervised by teachers. The handwriting and mathematical modules are experimentally validated, while the reading module is currently implemented as a prototype. Experimental results obtained from a stratified train/test split (80/20) and cross-validation indicated very high predictive power for dysgraphia (accuracy rate: 91%) and dyscalculia (accuracy rate: 94%). The reading component is available in the form of a prototype that has not been subjected to quantitative analysis yet; nevertheless, behavioral testing clearly indicated differences between normally developing children and children who had issues related to phonological processing and sequencing. Even though it cannot be considered a diagnostic tool, this system represents a practical and entertaining way of identifying children who might have learning problems, which may help parents and educators recognize when professional help should be sought out.
Naseer et al. (Wed,) studied this question.