Early screening of handwriting difficulties is key to start remediation activities that help distinguishing between a simple delay and dysgraphia. Technology is fundamental in this process, as also claimed by guidelines for dysgraphia diagnosis: it allows to implement artificial intelligence techniques to help in the discrimination of the difficulty. To this end, a serious game was leveraged to assess handwriting laws altered in dysgraphia starting from symbols drawing. 66 first and second graders were longitudinally tested both with the serious game and with a handwriting proficiency test. Objective features computed from the game were tested to understand if they significantly differed between children at risk and not at risk of dysgraphia, according to a standardized clinical test used to assess handwriting. Then, machine learning models were leveraged to predict the risk and understand the areas of difficulty. On average, 62% of the features significantly differ between risk levels for first graders, whilst only 35% for second graders, thus revealing a better sensitivity in younger children. This is encouraging for an early observation. As for machine learning, a Logistic classifier was able to predict risk with an area under the precision-recall curve of 0.84 for the risk class and 0.98 for the non-risk class. The results of this study could be a valid help for an artificial intelligence-enhanced screening of dysgraphia.

Serious games for a technology-enhanced early screening of handwriting difficulties

Fontolan, Stefania;Franceschini, Sandro;Bortolozzo, Marisa;Termine, Cristiano;
2023-01-01

Abstract

Early screening of handwriting difficulties is key to start remediation activities that help distinguishing between a simple delay and dysgraphia. Technology is fundamental in this process, as also claimed by guidelines for dysgraphia diagnosis: it allows to implement artificial intelligence techniques to help in the discrimination of the difficulty. To this end, a serious game was leveraged to assess handwriting laws altered in dysgraphia starting from symbols drawing. 66 first and second graders were longitudinally tested both with the serious game and with a handwriting proficiency test. Objective features computed from the game were tested to understand if they significantly differed between children at risk and not at risk of dysgraphia, according to a standardized clinical test used to assess handwriting. Then, machine learning models were leveraged to predict the risk and understand the areas of difficulty. On average, 62% of the features significantly differ between risk levels for first graders, whilst only 35% for second graders, thus revealing a better sensitivity in younger children. This is encouraging for an early observation. As for machine learning, a Logistic classifier was able to predict risk with an area under the precision-recall curve of 0.84 for the risk class and 0.98 for the non-risk class. The results of this study could be a valid help for an artificial intelligence-enhanced screening of dysgraphia.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11383/2179654
 Attenzione

L'Ateneo sottopone a validazione solo i file PDF allegati

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 0
social impact