Automated human-level diagnosis of dysgraphia using a consumer tablet

Area of Science:

• Pediatrics
• Neuroscience
• Educational Psychology

Background:

• Children's academic and behavioral progress depends on developing reading and writing skills.
• Dysgraphia, a handwriting disability, often co-occurs with neuro-developmental disorders like dyslexia and ADHD, impacting daily life and requiring intervention.
• Traditional handwriting assessments are subjective, costly, and lack scalability, failing to capture crucial motor control dynamics.

Purpose of the Study:

• To develop an objective, scalable, and cost-effective diagnostic tool for dysgraphia using commodity digital tablets.
• To leverage digital handwriting analysis to identify children with dysgraphia accurately.

Main Methods:

• Collected data from 298 children, including 56 diagnosed with dysgraphia, using a digital tablet for the BHK handwriting test.
• Extracted 53 distinct handwriting features, encompassing dynamics, pressure, and tilt, from the digital recordings.
• Employed a Random Forest classifier for dysgraphia diagnosis based on the extracted features.

Main Results:

• The developed tool achieved high diagnostic accuracy: 96.6% sensitivity and 99.2% specificity.
• The accuracy is comparable to expert assessments, considering inter-rater reliability.
• The method is suitable for direct deployment as a diagnostic tool.

Conclusions:

• A digital tablet-based approach offers a highly accurate and objective method for diagnosing dysgraphia.
• This technology enables early detection and intervention for handwriting difficulties.
• The tool provides a scalable and cost-effective solution for pediatric screening.