Differentiating ADHD subtypes and simulation of in silico therapeutic neuronal excitation in a whole-brain nonlinear dynamic modelling frame-work

Recent advances in non-linear computational and dynamical modelling have opened up the possibility to param etrize dynamic neural mechanisms that drive complex behavior. Importantly, building models of neuronal processes is of key importance to fully understand disorders of the brain as it may provide a quantitative platform that is capable of binding multiple neurophysiological processes to phenotype profiles. In this study, we apply a newly developed adaptive frequency-based model of whole-brain oscillations to resting-state fMRI data acquired from healthy controls and a cohort of attention deficit hyperactivity disorder (ADHD) subjects. As expected, we found that healthy control subjects differed from ADHD in terms of attractor dynamics. However, we also found a marked dichotomy in neural dynamics within the ADHD cohort. Next, we classified the ADHD group according to the level of distance of each individuals empirical network from the two model-based simulated networks. Critically, the model was mirrored in the empirical behavior data with the two ADHD subgroups displaying distinct behavioral phenotypes related to emotional instability (i.e., depression and hypomanic personality traits). Finally, we investigated the applicability and feasibility of our whole-brain model in a therapeutic setting by conducting in silico excitatory stimulations to parsimoniously mimic clinical neuro-stimulation paradigms in ADHD. We tested the effect of stimulating any individual brain region on the key network measures derived from the simulated brain network and its contribution in rectifying the brain dynamics to that of the healthy brain, separately for each ADHD subgroup. This showed that this was indeed possible for both subgroups. However, the current effect sizes were small suggesting that the stimulation protocol needs to be tailored at the individual level. These findings demonstrate the potential of this new modelling framework to unveil hidden neurophysiological profiles and establish tailored clinical interventions.