Exploring Functional Connectivity in Attention Deficit/Hyperactivity Disorder: A Functional Near-infrared Spectroscopy Study with Machine Learning Analysis.

Abstract

Functional near-infrared spectroscopy (fNIRS) has shown potential in attention deficit/hyperactivity disorder (ADHD) research, though it is not yet widely used as a primary diagnostic tool. While most previous studies have focused on children and resting-state conditions, research on adult ADHD, particularly under task-state conditions, is increasing but still limited compared to studies on children. Since ADHD is associated with cognitive challenges and alterations in brain activity, investigating functional connectivity during a task can provide a better understanding of its neural characteristics. In this study, we aim to investigate functional connectivity in adult patients with ADHD by comparing them with healthy controls under task-state conditions. We used the fNIRS dataset, which comprised 75 healthy controls and 75 medication-naïve individuals with ADHD. The network characteristics of functional connectivity were compared during a verbal fluency task, specifically focusing on density, global clustering coefficient, efficiency, and average betweenness centrality. By statistical analysis between the two groups, statistical significance was observed in density (p<0.001, t = 5.39, η2 = 0.443). Additionally, various machine learning classifiers were employed to assess the potential of functional connectivity metrics in classifying the two groups. The linear support vector machine achieved accuracy and precision of 0.800, recall of 0.808, and F1-score of 0.799, representing the highest performance among five different classifiers. In conclusion, our findings reveal distinct functional connectivity patterns among the groups, highlighting the potential of fNIRS-derived functional connectivity metrics as biomarkers for ADHD.