Early Diagnosis of Dysgraphia in Children Based on Handwriting Image Using AI

Dysgraphia is a neurological learning disability that hampers handwriting skills in children and often remains undiagnosed due to limited awareness and the absence of scalable diagnostic tools. This study proposes an AI-driven framework for the early detection of dysgraphia using handwriting images. To address the scarcity of real-world dysgraphic data, the framework introduces, to the best of current knowledge, the first synthetic handwriting dataset of dysgraphic samples generated with Generative Adversarial Networks (GANs) trained on labeled handwriting data. A hybrid feature extraction pipeline combining Optical Character Recognition (OCR) and Large Language Models (LLMs) is employed to capture both visual and linguistic cues from handwriting. These features are used to train a suite of machine learning classifiers, achieving a maximum accuracy of 85% in detecting handwriting indicative of dysgraphia. A web-based diagnostic platform that offers easily accessible, clear, and understandable handwriting analysis has been created to make real-world implementation easier. By providing a scalable and dependable early screening solution, the proposed approach not only enhances automated dysgraphia identification by enhancing interpretability and data availability but also supports therapeutic and educational interventions.