Predicting Outcomes From tDCS Intervention in Parkinson' Disease Using Electroencephalographic Biomarkers and Machine Learning Approach: the PREDICT Study Protocol

Parkinson's disease (PD) is a progressive and disabling neurodegenerative disease, clinically characterized by motor and non-motor symptoms. The potential of the "Transcranial direct current stimulation" (tDCS) for symptomatic improvement in these patients has been demonstrated, but the factors associated with the best therapeutic response are not known. The electroencephalogram (EEG) is considered as a diagnostic and prognostic biomarker of PD, and has been used in recent studies associated with machine-learning methods to identify predictors of responses in neurological and psychiatric conditions. Using connectivity-based prediction and machine-learning, the investigators intend to identify and compare characteristics related to baseline resting EEG between PD responders and non-responders to tDCS treatment. The recruited participants will be randomized to treatment with active tDCS associated with dual-task motor therapy or motor therapy with visual cues. A resting-state electroencephalography (EEG) will be recorded prior to the start of the treatment. The investigators will determine clinical improvement labels used for machine learning classification, in baseline and posttreatment assessments and will use three different methods to categorize the data into two classes (low or high improvement): Support Vector Machine (SVM), Linear Discriminant Analysis (LDA) and Extreme Learning Machine (ELM). The functional label will be based on the Timed Up and Go Test recorded at baseline and posttreament of tDCS treatment.