Mapping neural effects of mindfulness-based cognitive therapy in ADHD using EEG microstates and machine learning models.

Abstract

Mindfulness-based cognitive therapy (MBCT) is one of the promising treatments with no known side effects for neuropsychiatric conditions such as Attention-deficit/hyperactivity disorder (ADHD). However, the mechanism of action underlying MBCT is not clearly understood. Here, we applied resting-state EEG microstate analysis and machine learning modeling to characterize brain network dynamics in adults with ADHD exposed to MBCT. Sixty-one participants were randomized to a 12-week MBCT intervention or waitlist control (WL), with clinical assessments and EEG recordings collected pre-to-post trial. We analyzed the microstate dynamics of EEG data in different frequency bands, comparing four microstate classes (A-D), and the cross-correlation of microstate dynamics with clinical measures. Furthermore, machine learning computational techniques were applied to predict which patients can benefit more from the MBCT intervention based on their brain dynamics pre-treatment. Microstate analyses revealed significant MBCT-related alterations in temporal dynamics, including increased coverage and duration of microstates A and B, as well as changes in individual explained variance in microstate A (theta band) and microstate D (alpha band). Coverage and explained variance for microstate B also showed significant changes across the full spectrum. These changes were strongly correlated with improvements in ADHD symptomatology, mindfulness skills, quality of life, and executive function across seven clinical domains. Critically, machine learning models predicted individual treatment responses with 83% accuracy using microstate dynamics. These findings demonstrate that MBCT systematically reshapes resting-state neural microstates in ADHD, including microstate classes A, B, and D, and suggest that computational EEG biomarkers may inform precision approaches to mindfulness-based interventions.