Linear and non-linear autoregressive modeling in subthalamic nucleus for patients with movement disorders. Comparison and critical analysis

The Deep Brain Stimulation is a surgical procedure in which an electrode is implanted, used by functional neurosurgeons to control the discharge rates of motor units (basal ganglia) in patients with movement disorders. The success of the procedure depends on exactly localization of surgical target, conventionally the subthalamic nucleus, thalamus or globus pallidus internus, which have a particular voltage profile. In this work, two kind of parametric structures, non-linear ARX and linear AR, have been used for modeling the intracerebral signals in patients with Parkinson disease. This work evaluates the fitness with both modeling techniques and their dependence of the linearity regressors and the prediction horizon. The author found that the signals without Gaussian behavior were strongly sensitive of the prediction horizon. On the other hand, both AR and NLARX had good enough precision that guarantees an accurate simulation. This work aims to establish the better modeling criteria trough an a comparison between fitness for AR and NLARX structures and the final model of subthalamic nucleus signals for an oblique coordinate system.