Using Machine Learning To Identify Neural Mechanisms Underlying the Development of Cognition in Children and Adolescents With ADHD
Author: Brian Pho
Supervisor: Stojanoski, Bobby, Ontario Tech University, The University of Western Ontario
Co-Supervisor: Mohsenzadeh, Yalda, The University of Western Ontario
One of the most common disorders in children that affects the development of the brain is attention deficit hyperactivity disorder (ADHD). ADHD is defined by abnormal amounts of inattention, impulsivity, and hyperactivity—such as when a student has trouble paying attention to the teacher’s school lesson. From previous scientific studies, it is believed that ADHD is linked to problems with controlling a person’s behavior including suppressing inappropriate or unwanted actions, switching between different mental processes, and remembering information for a short period of time. However, it is not clear how these mental or cognitive abilities relate to children’s brains with ADHD and how they change as children become adolescents. Specifically, how does the brain activity in children and adolescents diagnosed with ADHD relate to cognitive abilities such as IQ and working memory? In this study, I explored this question by using computer models to find links between brain activity—while participants watched the movie “Despicable Me”—to various cognitive abilities in children and adolescents ages 6 to 16 with ADHD. I found that the models could predict IQ, visual spatial ability, and verbal comprehension ability in early childhood. In addition to these three cognitive abilities, the models could also predict fluid reasoning and working memory ability in middle childhood. This suggests that the models can capture different cognitive abilities in children of different ages. But how do the models predict cognition from brain activity? By analyzing the models, they pointed to a set of brain networks that were more important than other networks to predict cognitive ability. These networks are believed to participate in memory, attention, and motor control, and this set of important networks changed as the models were trained on children in different developmental stages. Overall, these findings provide evidence that computer models can predict cognitive ability in children with ADHD and they contribute to our understanding of how brain activity is linked to different cognitive abilities in children with ADHD.
This repository contains the code for my Masters in Neuroscience research project. The project started in September 2020 and finished in September 2022 under the supervision of Bobby Stojanoski and Yalda Mohsenzadeh at Western University (The University of Western Ontario).
To view this project's detailed documentation and results, head over to the wiki.
To read the completed thesis, download the PDF from here.