Dysgraphia Detection in Children Using Handwriting Samples

Dysgraphia, a neurological disorder that hinders handwriting abilities, can provide significant challenges for children if not identified early. This research offers a novel CNN-BiLSTM architecture for detecting dysgraphia in handwriting samples as part of the early diagnosis process. Our model uses CNNs to extract spatial features and BiLSTMs to capture sequential dependencies in handwriting patterns, resulting in a classification accuracy of 95%, equivalent to traditional CNN-based approaches. The proposed architecture provides a reliable diagnostic tool with minimal pre-processing, which makes it suitable for practical application. Previous works rely on CNN or shallow classifiers like SVM or RF while our hybrid CNN-BiLSTM exploits both spatial and sequential dependencies in handwriting images, which are not much explored in traditional dysgraphia detection. This study aims to provide educators, parents, and healthcare professionals with a simple and effective method to identify dysgraphia, resulting in timely interventions and inclusive education.