Abstract:Impaired motor vigor is a defining characteristic of Parkinson’s disease (PD), yet the underlying brain mechanisms governing motor vigor (MV) remain unclear. Recent studies have suggested beta-gamma Phase-Amplitude Coupling (PAC) derived from the resting-state electroencephalogram (EEG) is a potential biomarker for PD that is modulated by Deep Brain Stimulation (DBS) and L-dopa treatment. Specifically, PAC has been suggested to be a marker of transitions between motor movements, as opposed to encoding the vigor of the current movement. Here, we comprehensively investigate the potential of various PAC interactions—across different frequency pairs—beyond the linear approaches typically employed to predict MV during motor tasks in PD and study the effects of dopaminergic medication and non-invasive Galvanic Vestibular Stimulation (GVS). We recorded EEG data from 20 PD patients and 22 healthy controls executing a simple, overlearned handgrip task. Subjects were tested on and off L-dopa medication and with and without GVS (multi-sine either 50-100Hz, 100-150Hz). In a preliminary linear (LASSO-based) analysis comparing various PACs and a broad range of commonly used EEG features, PAC features were found to be crucial for predicting MV approximately equally in PD and HC. Initial findings from the linear analysis showed PAC as a significant indicator for MV in both groups, although with variability in cross-validation that implied a complex, non-linear relationship between PAC and MV. To extensively investigate the PAC-MV relation, we used a deep convolutional neural network (PACNET)—developed based on pre-trained VGG-16 architecture—to estimate MV from PAC values. In both PD and HCs, delta-beta, theta-, alpha-, and beta-gamma PACs were important for MV prediction. In PD subjects, GVS affected delta-beta, theta-gamma-, and beta-gamma PACs role in MV prediction, which was sensitive to different GVS stimulation parameters. These PACs were also relevant for PD patients’ MV prediction after L-dopa medication. This study supports the hypothesis that EEG PAC across multiple frequency pairs, not just beta-gamma, predicts MV and not just motor transitions and can be a biomarker for assessing the impact of electrical stimulation and dopaminergic medication in PD. Our results suggest that PAC is involved in MV, in addition to a range of previously reported cognitive processes, including working and long-term memory, attention, language, and fluid intelligence. Non-linear approaches appear important for examining EEG PAC and behavior relations.

Retrovirus-mediated Transfer of Nuclear Factor-KB Subunit Genes Modulates licBu and Interferon 3 Expression'

Disrupted morphological grey matter networks in early-stage Parkinson’s disease

A convolutional-recurrent neural network approach to resting-state EEG classification in Parkinson’s disease

Parkinson's disease detection based on multi-pattern analysis and multi-scale convolutional neural networks

Deep learning reveals personalized spatial spectral abnormalities of high delta and low alpha bands in EEG of patients with early Parkinson’s disease

Effect of angiotensin-converting enzyme inhibition on plasma brain natriuretic peptide levels in patients with heart failure.

Parkinson’s Disease Classification Using Machine Learning Approaches and Resting-State EEG

Parkinson’s Disease Detection from Resting-State EEG Signals Using Common Spatial Pattern, Entropy, and Machine Learning Techniques

Individual cases of Parkinson's disease can be robustly classified by cortical oscillatory activity from magnetoencephalography

Machine Learning-Based Detection of Parkinson's Disease From Resting-State EEG: A Multi-Center Study

Parkinsons Disease Detection via Resting-State Electroencephalography Using Signal Processing and Machine Learning Techniques

Fractal dimensions and machine learning for detection of Parkinson’s disease in resting-state electroencephalography

Oscillation-specific Nodal Alterations in Early to Middle Stages Parkinson's Disease

A Hybrid Deep Spatiotemporal Attention‐Based Model for Parkinson's Disease Diagnosis Using Resting State EEG Signals

On synthesis of 3 × 3 reversible logic functions

A Hybrid Deep Spatio-Temporal Attention-Based Model for Parkinson's Disease Diagnosis Using Resting State EEG Signals

Dynamical system based compact deep hybrid network for classification of Parkinson disease related EEG signals

Parkinson’s disease: deep learning with a parameter-weighted structural connectome matrix for diagnosis and neural circuit disorder investigation

An efficient Parkinson's disease detection framework: Leveraging time-frequency representation and AlexNet convolutional neural network

Survival in patients with oral and maxillofacial diffuse large B-cell lymphoma.

Predicting Motor Vigor from EEG Phase Amplitude Coupling (PAC) in Parkinson’s Disease: Effects of Dopaminergic Medication and Non-invasive Modulation