EEG Recordings and Analysis in Parkinson's Patients: Towards Adaptive Deep Brain Stimulation by Machine Learning

Overview

The objective of this protocol is to obtain on Parkinson's disease more accessible therapeutic targets than deep brain stimulation (HFS-STN), the neurosurgical treatment for this pathology. This study will pave the way for new forms of adaptive processing for the HFS-STN. It could become functionally coupled to a minimalist EEG centred on the motor cortex and to software for decoding, live or slightly delayed, classes of movements performed. On the one hand, this device could be used as a sensor of the quality of the information transmitted by the cortical network, thus allowing the selection of the optimal parameters of the HFS-STN on the basis of the movement decoding score. On the other hand, this device could lead to adapting the HFS-STN treatment over time by regularly calculating the recognition scores of the different movements performed and comparing them to the initial scores.

One of the therapies for Parkinson's disease, a condition affecting nearly 150,000 patients in France, is the invasive neurosurgical implantation of high-frequency deep brain stimulation of the subthalamic nuclei (HFS-STN). Although HFS-STN is very effective, the underlying mechanisms are still relatively poorly understood, particularly at the cortical level, a region that could become an alternative therapeutic target because it is easier to access. This study aims to measure the changes induced by the antiparkinsonian drug treatment and the HFS-STN on the encoding and transmission of motor information at the level of the motor cortex, thanks to the recording of the electroencephalogram of patients. These recordings, made during the performance of certain movements, will be subjected to an analysis using "machine learning" methods that will make it possible to decode the identity of the movement performed more or less efficiently.