Computational neurostimulation for Parkinson's disease

Deep brain stimulation (DBS) has had a remarkable success in treating a range of neurological and psychiatric conditions. However, efficacy remains suboptimal and patients can often develop side effects. The underlying causes of both the beneficial and detrimental effects of DBS remain incompletely understood which is delaying improvements to current DBS therapies and limiting developments of future treatments. Advancing this mechanistic understanding will require the design of appropriate models that can formalize the interaction between DBS and the cortico-basal-ganglia network. Recent advances in biophysical modeling have provided important insights into the impact of stimulation at local (neuronal membranes, electrical fields), intermediate (neural networks), and higher (phase, synchronization) levels of description. These have made important contributions to explaining neurophysiological changes during DBS (e.g., spikes, local field potentials), but such models generally do not seek to make accurate predictions about the resultant consequences on behavior. We argue that further advance will rest on models that focus on the specific computations that are performed in cortico-basal-ganglia networks, and address how DBS alters these computations and how this in turn modifies behavior. For the emergent field of computational modeling as applied to Parkinson's disease, we propose that models at mesoscopic levels of description are likely to be most valuable, with a particular focus on the role of oscillations and their relationship to behavior. It is therefore hoped that computational neurostimulation will usher in a new era of rapid, rationally derived DBS advancements for neurological and psychiatric disorders.