Main description (English)

The diagnosis of Learning Disabilities (LD) is frequently subject to cognitive biases. In Italy, minimal diagnostic standards have been identified during a national Consensus Conference (2010). However, specialists use different protocols to assess reading and cognitive abilities. Thus, we propose to support LDs diagnosis with Artificial Neural Networks (ANN). Clinical results from 203 reports were input to investigate which ones can predict LD diagnosis. In addition, correlations among LDs were explored. Preliminary results show that ANNs can be useful to support a clinical diagnosis of LDs with an 81.93% average accuracy, and, under certain conditions, with a 99% certainty. Additionally, the 10 most meaningful tests for each LD have been identified and significant correlations between dyscalculia and dyslexia were found.