Predicting ADHD from resting-state fMRI: "connectivity patterns using machine learning"

Abstract

[eng] Attention deficit hyperactivity disorder (ADHD) is a prevalent neurodevelopmental disorder affecting millions of children and adolescents. Despite extensive research, diagnosing ADHD remains challenging due to its multifaceted nature. This study investigates the potential of machine learning to classify ADHD using functional connectivity patterns derived from resting-state fMRI (rs-fMRI) data. We use Support Vector Machines (SVMs) to analyze connectivity features extracted from the ADHD-200 dataset, a publicly available collection of rs-fMRI data from individuals with ADHD and typically developing controls. We tried different non linear algorithms to improve the mode performance and overcome the moderate accuracy of the linear models in differentiating individuals with ADHD from controls, highlighting the potential of this approach for computational diagnosis. However, further research with a larger dataset is needed to improve classification performance.