Brain connectome differences between attention deficit hyperactivity disorder (ADHD) and neurotypical children during visual attention: A study using a minimum spanning tree graph, multichannel EEG recording and machine learning.

Background: Individuals with attention deficit hyperactivity disorder (ADHD) perform visual attention tasks differently compared to neurotypicals. In this study, differences of brain connectome during visual attention were compared between ADHD and neurotypicals using multichannel electrocardiogram (EEG) recordings and graph theory.

Methods: A minimum spanning tree (MST) graph based on similarities in EEG data from different brain areas was constructed for both neurotypical and ADHD groups. Features of MST were extracted in different EEG frequency sub-bands. The discriminative capability of MST extracted features was assessed using a classification approach. By comparing graph features between ADHD and neurotypicals, differences between brain processing mechanisms were investigated.

Results: Features extracted from the MST graph achieved a perfect discrimination between individuals with ADHD and neurotypicals (accuracy = 100%, AUC = 1). This result was consistent across multiple classifiers and different types of similarity measures used for graph construction. The most discriminative MST graph features were identified in the alpha band. Significantly reduced leaf number, mean eccentricity, radius, and diameter in the high alpha were the main results. Furthermore, the results revealed lack of frontal processing hubs and weaker frontoparietal connection in the ADHD group.

Conclusion: The results of this study indicated that MST graph features were ideal candidates for investigating underlying mechanisms of ADHD.