Advanced Computational Techniques for Dysgraphia Prediction through Handwriting Recognition Using Machine Learning and Deep Learning Methods

This systematic review critically examines the utilization of machine learning (ML) and deep learning (DL) techniques in detecting and predicting Dysgraphia through handwriting recognition. The primary research objective is to systematically assess the significance, contributions, performance, and limitations of recent technological advances in identifying dysgraphic patients. Employing a structured search and selection methodology based on the PRISMA protocol, 34 peer-reviewed studies from key academic databases, including IEEE Xplore, PubMed, Scopus, Google Scholar, and Science Direct, were comprehensively analyzed. The review methodology focused extensively on data collection methods, ethical considerations, preprocessing approaches, feature extraction and selection techniques, and the performance metrics of various ML models in dysgraphia prediction. Our key findings indicate that a Hybrid AI method combining convolutional neural networks and support vector machines demonstrates high accuracy rates, reaching up to 99.33%. This systematic review compares machine learning and deep learning use in diagnosing Dysgraphia through handwriting recognition with traditional cognitive evaluations, highlighting the latter's lack of scalability and high labour demands. Additionally, this review highlights that research on Dysgraphia has mainly focused on Latin scripts, overlooking the unique challenges of non-Latin scripts. It stresses the importance of including diverse linguistic data for more accurate global diagnoses. Additionally, Dysgraphia is influenced by cognitive, motor, and language skills, underscoring the need for multimodal data to enhance early detection and treatment effectiveness. The paper advocates for research that integrates various data types and expands linguistic diversity in training datasets, aiming to improve the diagnostic accuracy and global applicability of dysgraphia detection tools.