The impact of in-air features on the diagnosis of developmental dysgraphia.

This article focuses on the development and evaluation of an automated method for diagnosing developmental dysgraphia using online handwriting features, particularly emphasizing the role of in-air pen movements. The study collected handwriting data from 101 elementary students (49 dysgraphic and 52 non-dysgraphic) using a Wacom pen tablet capable of recording both on-surface and in-air pen kinematics, including position, pressure, pen angles, and temporal features. Machine learning classifiers—Support Vector Machine (SVM), Random Forest, and AdaBoost—were trained on extracted features, with the SVM achieving the highest accuracy of 85.7% when combining in-air and on-surface data, compared to 80.9% and 71.4% accuracy using only in-air or on-surface features, respectively. The findings highlight that in-air handwriting features contain significant diagnostic information and suggest that such automated, tablet-based assessments could serve as efficient, accessible tools to assist clinicians in identifying dysgraphia.