Multimodal MRI-based neuromarkers trace longitudinal changes in cognitive functioning in ADHD

Neuroimaging markers that capture interindividual differences may aid diagnosis and patient stratification, whereas those sensitive to intraindividual variation may inform prognosis and treatment monitoring. Yet, most machine-learning research in ADHD has emphasised interindividual prediction, leaving open whether neuroimaging can predict within-person cognitive changes in ADHD. Using a longitudinal ADHD-enriched dataset, the Oregon ADHD-1000 study (n = 594; 1,053 annual timepoints), we developed multimodal markers from resting-state functional and structural MRI to predict cognitive functioning. These markers showed good out-of-sample accuracy (r =.459) and performed comparably in children with and without ADHD. They explained 29.09% of interindividual variance in cognitive functioning, 33.48% of intraindividual cognitive trajectories, and 60.87% of intraindividual age-related cognitive development. They also accounted for substantial portions of the association between cognitive functioning and hyperactivity (58.79%) and inattention (25.99%). Together, these findings provide foundational evidence that f/sMRI-based markers can longitudinally track cognitive functioning, underscoring their potential, once further developed, for prognosis and treatment monitoring.