Scientists use machine learning to better predict learning disabilities in children



Scientists who use machine learning – a kind of artificial intelligence – with data from hundreds of children struggling at school, identified clusters of learning disabilities that did not match the earlier diagnosis that the children had received.

The researchers from the Cognition and Brain Sciences Unit of the Medical Research Council (MRC) of the University of Cambridge say that this requires the need for children to get detailed assessments of their cognitive skills to identify the best type of support.

The study, published in Developmental science, recruited 550 children who were referred to a clinic – the Center for Attention Learning and Memory – because they were struggling at school.

The scientists say that much of the previous research on learning disabilities focused on children who had already received a specific diagnosis, such as attention deficit hyperactivity disorder (ADHD), autism spectrum disorder or dyslexia. By including children with all difficulties, regardless of diagnosis, this study could better determine the range of problems within and the overlap between the diagnostic categories.

Dr. Duncan Astle of the MRC Cognition and Brain Sciences Unit at the University of Cambridge, who led the study, said: "Receiving a diagnosis is an important point of recognition for parents and children with learning disabilities who recognize the child's difficulties and help them Access to support But parents and professionals who work with these children on a daily basis see that neat labels do not record their individual problems – for example, the ADHD of a child is often not like another person's ADHD.

"Our study is the first of its kind to apply machine learning to a broad spectrum of hundreds of struggling students."

The team did this by providing the computer algorithm with a lot of cognitive test data from each child, including measurements of listening comprehension, spatial reasoning, problem solving, vocabulary and memory. Based on these data, the algorithm suggests that the children best fit into four clusters of difficulties.

These clusters closely match other data about the children, such as parents' reports about their communication problems and educational data about reading and arithmetic. But there was no correspondence with their previous diagnoses. To check whether these groups corresponded to biological differences, the groups were checked for MRI brain scans of 184 children. The groupings reflected patterns in connectivity within parts of the children's brains, suggesting that machine learning identified differences that partially reflect the underlying biology.

Two of the four identified groups were: problems with working with memory, and problems with processing sounds in words.

Difficulties with working memory – the storage and manipulation of information in the short term – have been associated with wrestling with mathematics and with tasks such as following lists. Difficulties in processing the sounds in words, called phonological skills, have been associated with struggling with reading.

Dr. Astle said: "Research in the past that has been selected for children with poor reading ability has shown a close link between struggling with reading and problems with word processing in words, but by looking at children with a wide range of problems, discovered we unexpectedly that many children with problems with the processing of sounds in words not only have problems reading – they also have problems with mathematics.

"If researchers study learning disabilities, we need to go beyond the diagnostic label and we hope that this study will help to develop better interventions that are more specifically focused on the individual cognitive problems of children."

Dr Joni Holmes, of the MRC Cognition and Brain Sciences Unit at the University of Cambridge, who was senior author of the study, said: "Our work suggests that children who find the same subjects difficult, might struggle for very different reasons, which has important implications. for selecting suitable interventions. "

The other two identified clusters were: children with broad cognitive problems in many areas and children with typical cognitive test results for their age. The researchers noted that the children in the group with cognitive test results that were typical of their age still had other problems that affected their schooling, such as behavioral problems that were not included in machine learning.

Dr. Joanna Latimer, Head of Neuroscience and Mental Health at the MRC, said: "These are interesting, early findings that begin to explore how we can apply new technologies, such as machine learning, to better understand brain function. in the brain to help develop better ways to support children with learning disabilities. "


Source link