Figure 1. Schematic depiction of general aDBS framework.

General closed loop DBS for adaptive adjustment of deep brain stimulation (DBS) parameters based upon real time patient measurements, such as electrophysiological signals (LFP, M/EEG, EMG), neurochemical parameters and behavioural measurements and machine learning. First, features from different possible signal sources are learned (e.g. beta frequency amplitude, phase of tremor oscillations) using deep learning approaches to classify between different behavioural (clinical) states (e.g. bradykinesia, tremor) and corresponding neural states. Then, actual states are compared with ideal states and stimulation parameters are adjusted and finally learned via reinforcement learning. In this closed-loop scheme, the stimulation parameters are adjusted within clinical limits based upon the difference between actual neural/behavioural and desired neural/behavioural state.

aDBS: adaptive deep brain stimulation; EEG: electroencephalography; EMG: electromyography; LFP: local field potentials