EEG-estimated functional connectivity, and not behavior, differentiates Parkinson's patients from health controls during the Simon conflict task

TL;DR

This work investigates EEG-estimated functional connectivity (FC) as a Parkinson’s Disease (PD) biomarker and finds that FC, estimated from delta and theta oscillations, yields spatial FC patterns significantly better at distinguishing PD from HC than temporal features or behavior.

Abstract

Neural biomarkers that can classify or predict disease are of broad interest to the neurological and psychiatric communities. Such biomarkers can be informative of disease state or treatment efficacy, even before there are changes in symptoms and/or behavior. This work investigates EEG-estimated functional connectivity (FC) as a Parkinson's Disease (PD) biomarker. Specifically, we investigate FC mediated via neural oscillations and consider such activity during the Simons conflict task. This task yields sensory-motor conflict, and one might expect differences in behavior between PD patients and healthy controls (HCs). In addition to considering spatially focused approaches, such as FC, as a biomarker, we also consider temporal biomarkers, which are more sensitive to ongoing changes in neural activity. We find that FC, estimated from delta (1-4Hz) and theta (4-7Hz) oscillations, yields spatial FC patterns significantly better at distinguishing PD from HC than temporal features or behavior. This study reinforces that FC in spectral bands is informative of differences in brain-wide processes and can serve as a biomarker distinguishing normal brain function from that seen in disease.

Figures (4)

• Figure 1: Overview of the experimental setup and the extracted temporal/spatial features. a. Illustration of the Simon conflict task (see text for details). b.1 Epoched 1-second temporal feature after the stimulus onset, specifically the voltage at electrode Cz within the delta band for incongruent trials. The average response time (RT) is $549.6\pm113.1$ ms for HCs and $563.2\pm156.2$ ms for PD patients in the incongruent trials (highlighted with the vertical dashed line separately, see Fig. \ref{['fig:binary']}a.1 and a.2 for more details about RT). b.2 A corresponding spatial feature from the same time window. The difference in functional connectivity (represented by Pearson Correlation) between HC and PD patients is plotted as an example.
• Figure 2: Prediction accuracy (HC vs. PD) from ML models. a.1 Model performance with features in the delta band (1-4Hz). Among the LSTM and CNN models trained and evaluated (see Table \ref{['tab:ML']}), we show the model having the highest test accuracy (highlighted with square markers). The bar plots are generated using 10-fold cross-validation results (mean $\pm$ standard deviation). Each model's highest validation accuracy is highlighted with a circle marker. We used the Mann-Whitney-Wilcoxon test given the small sample sizes (n=10) fay2010wilcoxon. The significance level of the test result (multi-comparison un-corrected) is indicated by a black asterisk, where *** indicates significance under a 99.9% confidence interval, ** indicates significance under a 99% confidence interval, and * indicates significance under a 95% confidence interval. a.2 is similar to a.1, except that models are evaluated in the theta band (4-7Hz).
• Figure 3: Classification probability given different biomarkers. a.1 Histogram of RT and receiver operating characteristic (ROC) curve of HC vs PD classification based on reaction time (RT). a.2 is similar to a.1, but for the incongruent trials. b.1 Subject-level probabilities of classifying HC vs PD. Both spatial FC (yellow) and temporal BPT (green) biomarkers outperform RT (gray), with 26 controls having probabilities less than 0.5 and 49 patients having probabilities greater than 0.5. The spatial FC biomarkers are clearly more accurate than the temporal BIT biomarkers across the population.
• Figure 4: The surrogate test identified significantly ($p<0.05$) different functional connectivity (FC) patterns under different thresholds (60% and 80%). a.1 Different FC patterns between groups within the delta band. Each color represents a comparison within one trial condition (red: congruent, labeled as 1 in the legend, e.g., C1 means HC-congruent; blue: incongruent, labeled as 2 in the legend, purple: shared in both conditions). a.2 follows the layout of a.1, except that it is computed based on the theta band.